Publish Kokoro PL/DE model catalog v1.0.0

.gitattributes

@@ -1,35 +1,3 @@
-*.7z filter=lfs diff=lfs merge=lfs -text
-*.arrow filter=lfs diff=lfs merge=lfs -text
 *.bin filter=lfs diff=lfs merge=lfs -text
-*.bz2 filter=lfs diff=lfs merge=lfs -text
-*.ckpt filter=lfs diff=lfs merge=lfs -text
-*.ftz filter=lfs diff=lfs merge=lfs -text
-*.gz filter=lfs diff=lfs merge=lfs -text
-*.h5 filter=lfs diff=lfs merge=lfs -text
-*.joblib filter=lfs diff=lfs merge=lfs -text
-*.lfs.* filter=lfs diff=lfs merge=lfs -text
-*.mlmodel filter=lfs diff=lfs merge=lfs -text
-*.model filter=lfs diff=lfs merge=lfs -text
-*.msgpack filter=lfs diff=lfs merge=lfs -text
-*.npy filter=lfs diff=lfs merge=lfs -text
-*.npz filter=lfs diff=lfs merge=lfs -text
 *.onnx filter=lfs diff=lfs merge=lfs -text
-*.ot filter=lfs diff=lfs merge=lfs -text
-*.parquet filter=lfs diff=lfs merge=lfs -text
-*.pb filter=lfs diff=lfs merge=lfs -text
-*.pickle filter=lfs diff=lfs merge=lfs -text
-*.pkl filter=lfs diff=lfs merge=lfs -text
-*.pt filter=lfs diff=lfs merge=lfs -text
-*.pth filter=lfs diff=lfs merge=lfs -text
-*.rar filter=lfs diff=lfs merge=lfs -text
-*.safetensors filter=lfs diff=lfs merge=lfs -text
-saved_model/**/* filter=lfs diff=lfs merge=lfs -text
-*.tar.* filter=lfs diff=lfs merge=lfs -text
-*.tar filter=lfs diff=lfs merge=lfs -text
-*.tflite filter=lfs diff=lfs merge=lfs -text
-*.tgz filter=lfs diff=lfs merge=lfs -text
-*.wasm filter=lfs diff=lfs merge=lfs -text
-*.xz filter=lfs diff=lfs merge=lfs -text
-*.zip filter=lfs diff=lfs merge=lfs -text
-*.zst filter=lfs diff=lfs merge=lfs -text
-*tfevents* filter=lfs diff=lfs merge=lfs -text
+*.wav filter=lfs diff=lfs merge=lfs -text

.gitignore

@@ -0,0 +1,5 @@
+.DS_Store
+:memory:.ses
+__pycache__/
+*.pyc
+node_modules/

LICENSE

@@ -0,0 +1,201 @@
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

README.md

@@ -0,0 +1,129 @@
+---
+license: apache-2.0
+pipeline_tag: text-to-speech
+library_name: onnxruntime
+language:
+  - pl
+  - de
+base_model: hexgrad/Kokoro-82M
+tags:
+  - kokoro
+  - onnx
+  - on-device
+  - kotlin-multiplatform
+  - android
+  - jvm
+  - web
+---
+
+# Kokoro PL/DE for ONNX Runtime and Kotlin Multiplatform
+
+Native Polish and German Kokoro-82M language packs prepared for on-device inference with ONNX
+Runtime. The repository is organized for a client that first chooses a language and then a voice.
+Each language can be downloaded independently; users do not need to fetch both packs.
+
+## Available packs
+
+| Language | Model | Voice | Model size | Phonemizer | Voice |
+|---|---|---|---:|---:|---:|
+| Polish (`pl-PL`) | `kokoro-pl-fp32` | Mateusz (male) | 310.5 MiB | 6.8 MiB | 510 KiB |
+| German (`de-DE`) | `kokoro-de-fp32` | Anna (female) | 310.5 MiB | 6.8 MiB | 510 KiB |
+
+The complete downloadable pack for one language is about 318 MiB plus tiny JSON metadata. Models
+are FP32 to preserve quality; no quantized variant is included in version 1.0.0.
+
+## Client flow
+
+1. Fetch [`catalog.json`](catalog.json) from a pinned Hub revision such as `v1.0.0`.
+2. Show `languages`, then show the selected language's `voices`.
+3. Download only that language's `model.artifact`, `phonemizer.artifact`, the selected voice
+   `artifact`, and the shared 2.3 KiB tokenizer config.
+4. Verify `sizeBytes` and `sha256` before activating a download.
+5. Resolve artifact paths relative to the catalog URL, so a tagged catalog automatically keeps all
+   files on the same immutable release.
+
+The JSON format is versioned and described by
+[`schema/kokoro-catalog.schema.json`](schema/kokoro-catalog.schema.json). New voices can be added to
+one language without changing the application API.
+
+## Inference contract
+
+The ONNX graph takes:
+
+- `input_ids`: `int64[1, sequence_length]`, with token `0` added at both ends;
+- `style`: `float32[1, 256]`;
+- `speed`: `float32[1]`.
+
+It returns a mono `float32` waveform at 24 kHz and an `int64` duration for every input token.
+Batch size is fixed to one; sequence length is dynamic up to 510 phoneme Unicode code points plus
+the two boundary tokens.
+
+Voice files are raw little-endian `float32[510, 256]` matrices. Select style row
+`phonemeCodePointCount - 1`. Count Unicode code points, not Kotlin/JVM UTF-16 code units; combining
+marks such as the nasalization mark in `ɔ̃` count separately.
+
+## Phonemization
+
+The included 6.8 MiB `.bin` files are neural Protophone data for
+[`Phonemis`](https://github.com/IgorSwat/Phonemis). They produce Kokoro-compatible IPA for Polish
+and German without shipping eSpeak or GPL components. Phonemis itself is a small C++20 library and
+is not embedded in these model-data files.
+
+Android/JVM consumers can expose Phonemis through a shared native bridge. A Web client needs a
+WebAssembly build of that bridge; this repository provides the same language data but does not ship
+a JavaScript/Wasm wrapper.
+
+## Validation
+
+Both graphs pass ONNX validation, ONNX Runtime 1.29 CPU inference, and real
+`onnxruntime-web 1.29.0` WebAssembly inference. Export validation compares each graph with its
+PyTorch checkpoint. Token durations match exactly; waveform correlation is `0.9948` for Polish and
+`0.9970` for German. Kokoro intentionally injects a small random excitation noise, so waveforms are
+not expected to be bit-identical across runs.
+
+Four natural-language runtime cases plus explicit 1-, 128-, and 510-phoneme boundary cases for
+both languages are recorded in
+[`validation/package-validation.json`](validation/package-validation.json). Generated previews are
+available as [`samples/pl-mateusz.wav`](samples/pl-mateusz.wav) and
+[`samples/de-anna.wav`](samples/de-anna.wav).
+
+To repeat the local checks after rebuilding the graphs:
+
+```bash
+python scripts/build_catalog.py
+python scripts/validate_package.py
+npm install
+npm run validate:web
+```
+
+Maintainers can publish the fully validated folder after authenticating with the Hugging Face CLI:
+
+```bash
+python scripts/publish_hf.py <namespace>/kokoro-kmp-models
+```
+
+The publisher excludes local dependencies and refuses to overwrite an existing immutable release
+tag with a different commit.
+
+## Sources and reproducibility
+
+All source repositories, immutable revisions, and source checksums are pinned in
+[`SOURCES.lock.json`](SOURCES.lock.json). `scripts/export_onnx.py` exports a self-contained opset 17
+graph and rejects outputs that fail ONNX validation or diverge materially from PyTorch.
+
+## Limitations
+
+- Version 1.0.0 contains one native voice per language.
+- Polish and German use separate fine-tuned model weights; a voice file cannot turn one language
+  model into the other.
+- Text normalization and pronunciation still need application-level tests for names,
+  abbreviations, mixed-language text, and unusual numbers.
+- The graphs are intentionally unquantized and prioritize quality over download size.
+
+## License
+
+Kokoro, the fine-tuned checkpoints, the converted ONNX graphs, speaker embeddings, and this
+packaging remain under **Apache License 2.0**. The repository license has not been changed.
+
+Phonemis assets remain under their upstream **MIT License**. See
+[`THIRD_PARTY_NOTICES.md`](THIRD_PARTY_NOTICES.md) for exact provenance and license boundaries.

SOURCES.lock.json

@@ -0,0 +1,31 @@
+{
+  "kokoroBase": {
+    "repository": "hexgrad/Kokoro-82M",
+    "revision": "f3ff3571791e39611d31c381e3a41a3af07b4987",
+    "files": {
+      "config.json": "5abb01e2403b072bf03d04fde160443e209d7a0dad49a423be15196b9b43c17f"
+    }
+  },
+  "languageModels": {
+    "repository": "software-mansion/react-native-executorch-kokoro",
+    "revision": "3744b57964eab7df6e8c48f0b84badb29e14df07",
+    "files": {
+      "finetunes/kokoro_polish_converted.pth": "e3202dc4d1f6e65dddff8a8e8d2e091ce9a3a66cf989c6c581a81b2f1969af49",
+      "finetunes/kokoro_german_converted.pth": "b8b2ab322963e7662c6036035c76c34a6a5582f814917407c94d632a1c930f71",
+      "voices/pm_mateusz.bin": "dc8f2919ede945e6962310b5204b912e64689d5698c571822bf75ad704870cb4",
+      "voices/df_anna.bin": "d583ccff3cdca2f7fae535cb998ac07e9fcb90f09737b9a41fa2734ec44a8f0b"
+    }
+  },
+  "phonemis": {
+    "repository": "IgorSwat/Phonemis",
+    "revision": "71eb1ce33bd586d38cbac037843b8539d7829c3b",
+    "files": {
+      "data/pl/phonemizer_pl.bin": "ec85f4dc2c4ac7a72ff88b98b0664a4ed887bd15c0d5add2eb1d6a6ee05b73f2",
+      "data/de/phonemizer_de.bin": "4888dc7e54dc66098551555096562063364091fc246da0d057b629587bedaa0b"
+    }
+  },
+  "exporter": {
+    "repository": "SuvioMedia/kokoro",
+    "revision": "8ee8ec3179c38990dd0564fb703c4deb68be58e9"
+  }
+}

THIRD_PARTY_NOTICES.md

@@ -0,0 +1,31 @@
+# Third-party notices
+
+This repository keeps the Kokoro licensing intact. Packaging and conversion do not replace or
+relicense upstream work.
+
+## Kokoro and the Polish/German fine-tunes
+
+- Base model and architecture: [`hexgrad/Kokoro-82M`](https://huggingface.co/hexgrad/Kokoro-82M),
+  revision `f3ff3571791e39611d31c381e3a41a3af07b4987`.
+- Polish and German converted checkpoints and speaker embeddings:
+  [`software-mansion/react-native-executorch-kokoro`](https://huggingface.co/software-mansion/react-native-executorch-kokoro),
+  revision `3744b57964eab7df6e8c48f0b84badb29e14df07`.
+- ONNX exporter: [`SuvioMedia/kokoro`](https://github.com/SuvioMedia/kokoro), revision
+  `8ee8ec3179c38990dd0564fb703c4deb68be58e9`.
+
+These materials and the resulting ONNX model files are distributed under Apache License 2.0.
+The license text is available in [`LICENSE`](LICENSE) and [`third-party/kokoro/LICENSE`](third-party/kokoro/LICENSE).
+
+## Phonemis
+
+The files below come from [`IgorSwat/Phonemis`](https://github.com/IgorSwat/Phonemis), revision
+`71eb1ce33bd586d38cbac037843b8539d7829c3b`:
+
+- `phonemizers/pl/phonemizer_pl.bin`
+- `phonemizers/de/phonemizer_de.bin`
+
+Phonemis is distributed under the MIT License. Its unchanged license text is available in
+[`third-party/phonemis/LICENSE`](third-party/phonemis/LICENSE). The MIT exception applies to
+the Phonemis files; it does not alter the Apache-2.0 license of Kokoro.
+
+Exact source file hashes are recorded in [`SOURCES.lock.json`](SOURCES.lock.json).

catalog.json

@@ -0,0 +1,184 @@
+{
+  "$schema": "schema/kokoro-catalog.schema.json",
+  "schemaVersion": 1,
+  "catalogVersion": "1.0.0",
+  "distribution": {
+    "provider": "huggingface",
+    "artifactResolution": "relative-to-catalog",
+    "recommendedRevision": "v1.0.0"
+  },
+  "license": "Apache-2.0",
+  "modelFamily": "kokoro-82m",
+  "audio": {
+    "sampleRateHz": 24000,
+    "channels": 1,
+    "sampleType": "float32"
+  },
+  "runtime": {
+    "engine": "onnxruntime",
+    "opset": 17,
+    "batchSize": 1,
+    "maxPhonemeCodePoints": 510,
+    "durationToSamplesFactor": 600,
+    "inputs": [
+      {
+        "name": "input_ids",
+        "type": "int64",
+        "shape": [
+          1,
+          "sequence_length"
+        ]
+      },
+      {
+        "name": "style",
+        "type": "float32",
+        "shape": [
+          1,
+          256
+        ]
+      },
+      {
+        "name": "speed",
+        "type": "float32",
+        "shape": [
+          1
+        ]
+      }
+    ],
+    "outputs": [
+      {
+        "name": "waveform",
+        "type": "float32",
+        "shape": [
+          "num_samples"
+        ]
+      },
+      {
+        "name": "duration",
+        "type": "int64",
+        "shape": [
+          "sequence_length"
+        ]
+      }
+    ],
+    "tokenizer": {
+      "bosTokenId": 0,
+      "eosTokenId": 0,
+      "vocabularyField": "vocab",
+      "artifact": {
+        "path": "runtime/config.json",
+        "mediaType": "application/json",
+        "sizeBytes": 2351,
+        "sha256": "5abb01e2403b072bf03d04fde160443e209d7a0dad49a423be15196b9b43c17f"
+      }
+    },
+    "voiceStyle": {
+      "format": "raw-float32-le",
+      "shape": [
+        510,
+        256
+      ],
+      "rowSelection": "phoneme_code_point_count_minus_one"
+    }
+  },
+  "languages": [
+    {
+      "id": "pl",
+      "locale": "pl-PL",
+      "displayName": "Polish",
+      "nativeName": "polski",
+      "model": {
+        "id": "kokoro-pl-fp32",
+        "format": "onnx",
+        "precision": "fp32",
+        "artifact": {
+          "path": "models/pl/kokoro-pl-fp32.onnx",
+          "mediaType": "application/onnx",
+          "sizeBytes": 325563781,
+          "sha256": "66637ea517e8ea4fea6c473c18a6f3b4fe982a8e7eb3879bb5f8250c7dc5679a"
+        },
+        "validation": {
+          "path": "models/pl/kokoro-pl-fp32.validation.json",
+          "mediaType": "application/json",
+          "sizeBytes": 1007,
+          "sha256": "63768a6dde6c8dabecec1ad673b3819c8d707f8e3cf565dda40950ec9c5c4476"
+        }
+      },
+      "phonemizer": {
+        "engine": "phonemis-protophone",
+        "language": "pl",
+        "license": "MIT",
+        "sourceRevision": "71eb1ce33bd586d38cbac037843b8539d7829c3b",
+        "artifact": {
+          "path": "phonemizers/pl/phonemizer_pl.bin",
+          "mediaType": "application/octet-stream",
+          "sizeBytes": 7094120,
+          "sha256": "ec85f4dc2c4ac7a72ff88b98b0664a4ed887bd15c0d5add2eb1d6a6ee05b73f2"
+        }
+      },
+      "voices": [
+        {
+          "id": "mateusz",
+          "displayName": "Mateusz",
+          "gender": "male",
+          "artifact": {
+            "path": "voices/pl/pm_mateusz.bin",
+            "mediaType": "application/octet-stream",
+            "sizeBytes": 522240,
+            "sha256": "dc8f2919ede945e6962310b5204b912e64689d5698c571822bf75ad704870cb4"
+          }
+        }
+      ],
+      "previewAudioPath": "samples/pl-mateusz.wav"
+    },
+    {
+      "id": "de",
+      "locale": "de-DE",
+      "displayName": "German",
+      "nativeName": "Deutsch",
+      "model": {
+        "id": "kokoro-de-fp32",
+        "format": "onnx",
+        "precision": "fp32",
+        "artifact": {
+          "path": "models/de/kokoro-de-fp32.onnx",
+          "mediaType": "application/onnx",
+          "sizeBytes": 325563781,
+          "sha256": "ecea710e0e4481d5930fc3486c5ed1c3dfd2dc752c2beb676135343d0236b1f3"
+        },
+        "validation": {
+          "path": "models/de/kokoro-de-fp32.validation.json",
+          "mediaType": "application/json",
+          "sizeBytes": 1142,
+          "sha256": "b9c7ff15e5be4824ff9e608d26ec21026d9e33ee3d4b430ad8ffecaa7af9e2d8"
+        }
+      },
+      "phonemizer": {
+        "engine": "phonemis-protophone",
+        "language": "de",
+        "license": "MIT",
+        "sourceRevision": "71eb1ce33bd586d38cbac037843b8539d7829c3b",
+        "artifact": {
+          "path": "phonemizers/de/phonemizer_de.bin",
+          "mediaType": "application/octet-stream",
+          "sizeBytes": 7094120,
+          "sha256": "4888dc7e54dc66098551555096562063364091fc246da0d057b629587bedaa0b"
+        }
+      },
+      "voices": [
+        {
+          "id": "anna",
+          "displayName": "Anna",
+          "gender": "female",
+          "artifact": {
+            "path": "voices/de/df_anna.bin",
+            "mediaType": "application/octet-stream",
+            "sizeBytes": 522240,
+            "sha256": "d583ccff3cdca2f7fae535cb998ac07e9fcb90f09737b9a41fa2734ec44a8f0b"
+          }
+        }
+      ],
+      "previewAudioPath": "samples/de-anna.wav"
+    }
+  ]
+}

catalog.template.json

@@ -0,0 +1,169 @@
+{
+  "$schema": "schema/kokoro-catalog.schema.json",
+  "schemaVersion": 1,
+  "catalogVersion": "1.0.0",
+  "distribution": {
+    "provider": "huggingface",
+    "artifactResolution": "relative-to-catalog",
+    "recommendedRevision": "v1.0.0"
+  },
+  "license": "Apache-2.0",
+  "modelFamily": "kokoro-82m",
+  "audio": {
+    "sampleRateHz": 24000,
+    "channels": 1,
+    "sampleType": "float32"
+  },
+  "runtime": {
+    "engine": "onnxruntime",
+    "opset": 17,
+    "batchSize": 1,
+    "maxPhonemeCodePoints": 510,
+    "durationToSamplesFactor": 600,
+    "inputs": [
+      {
+        "name": "input_ids",
+        "type": "int64",
+        "shape": [1, "sequence_length"]
+      },
+      {
+        "name": "style",
+        "type": "float32",
+        "shape": [1, 256]
+      },
+      {
+        "name": "speed",
+        "type": "float32",
+        "shape": [1]
+      }
+    ],
+    "outputs": [
+      {
+        "name": "waveform",
+        "type": "float32",
+        "shape": ["num_samples"]
+      },
+      {
+        "name": "duration",
+        "type": "int64",
+        "shape": ["sequence_length"]
+      }
+    ],
+    "tokenizer": {
+      "bosTokenId": 0,
+      "eosTokenId": 0,
+      "vocabularyField": "vocab",
+      "artifact": {
+        "$artifact": {
+          "path": "runtime/config.json",
+          "mediaType": "application/json"
+        }
+      }
+    },
+    "voiceStyle": {
+      "format": "raw-float32-le",
+      "shape": [510, 256],
+      "rowSelection": "phoneme_code_point_count_minus_one"
+    }
+  },
+  "languages": [
+    {
+      "id": "pl",
+      "locale": "pl-PL",
+      "displayName": "Polish",
+      "nativeName": "polski",
+      "model": {
+        "id": "kokoro-pl-fp32",
+        "format": "onnx",
+        "precision": "fp32",
+        "artifact": {
+          "$artifact": {
+            "path": "models/pl/kokoro-pl-fp32.onnx",
+            "mediaType": "application/onnx"
+          }
+        },
+        "validation": {
+          "$artifact": {
+            "path": "models/pl/kokoro-pl-fp32.validation.json",
+            "mediaType": "application/json"
+          }
+        }
+      },
+      "phonemizer": {
+        "engine": "phonemis-protophone",
+        "language": "pl",
+        "license": "MIT",
+        "sourceRevision": "71eb1ce33bd586d38cbac037843b8539d7829c3b",
+        "artifact": {
+          "$artifact": {
+            "path": "phonemizers/pl/phonemizer_pl.bin",
+            "mediaType": "application/octet-stream"
+          }
+        }
+      },
+      "voices": [
+        {
+          "id": "mateusz",
+          "displayName": "Mateusz",
+          "gender": "male",
+          "artifact": {
+            "$artifact": {
+              "path": "voices/pl/pm_mateusz.bin",
+              "mediaType": "application/octet-stream"
+            }
+          }
+        }
+      ],
+      "previewAudioPath": "samples/pl-mateusz.wav"
+    },
+    {
+      "id": "de",
+      "locale": "de-DE",
+      "displayName": "German",
+      "nativeName": "Deutsch",
+      "model": {
+        "id": "kokoro-de-fp32",
+        "format": "onnx",
+        "precision": "fp32",
+        "artifact": {
+          "$artifact": {
+            "path": "models/de/kokoro-de-fp32.onnx",
+            "mediaType": "application/onnx"
+          }
+        },
+        "validation": {
+          "$artifact": {
+            "path": "models/de/kokoro-de-fp32.validation.json",
+            "mediaType": "application/json"
+          }
+        }
+      },
+      "phonemizer": {
+        "engine": "phonemis-protophone",
+        "language": "de",
+        "license": "MIT",
+        "sourceRevision": "71eb1ce33bd586d38cbac037843b8539d7829c3b",
+        "artifact": {
+          "$artifact": {
+            "path": "phonemizers/de/phonemizer_de.bin",
+            "mediaType": "application/octet-stream"
+          }
+        }
+      },
+      "voices": [
+        {
+          "id": "anna",
+          "displayName": "Anna",
+          "gender": "female",
+          "artifact": {
+            "$artifact": {
+              "path": "voices/de/df_anna.bin",
+              "mediaType": "application/octet-stream"
+            }
+          }
+        }
+      ],
+      "previewAudioPath": "samples/de-anna.wav"
+    }
+  ]
+}

models/de/kokoro-de-fp32.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ecea710e0e4481d5930fc3486c5ed1c3dfd2dc752c2beb676135343d0236b1f3
+size 325563781

models/de/kokoro-de-fp32.validation.json

@@ -0,0 +1,23 @@
+{
+  "language": "de-DE",
+  "model": "kokoro-de-fp32.onnx",
+  "modelBytes": 325563781,
+  "modelSha256": "ecea710e0e4481d5930fc3486c5ed1c3dfd2dc752c2beb676135343d0236b1f3",
+  "sourceCheckpoint": "kokoro_german_converted.pth",
+  "sourceCheckpointSha256": "b8b2ab322963e7662c6036035c76c34a6a5582f814917407c94d632a1c930f71",
+  "sourceRevision": "3744b57964eab7df6e8c48f0b84badb29e14df07",
+  "baseRevision": "f3ff3571791e39611d31c381e3a41a3af07b4987",
+  "opset": 17,
+  "sampleRateHz": 24000,
+  "voice": "df_anna.bin",
+  "voiceSha256": "d583ccff3cdca2f7fae535cb998ac07e9fcb90f09737b9a41fa2734ec44a8f0b",
+  "referencePhonemes": "f\u02c8al\u0283\u0259s \u02c8y\u02d0b\u0259n f\u0254n ks\u02c8y\u02d0lo\u02d0f\u02cc\u0254nmu\u02d0z\u02cci\u02d0k kv\u02c8\u025blt j\u02c8e\u02d0d\u0259n \u0261\u027e\u02c8\u00f8\u02d0s\u0259r\u0259n \u02a6v\u02c8\u025b\u027ek.",
+  "referenceInputTokens": 75,
+  "referenceOutputSamples": 90600,
+  "maxAbsoluteError": 0.06611382961273193,
+  "meanAbsoluteError": 0.0024777022190392017,
+  "waveformCorrelation": 0.996961705255154,
+  "durationExactMatch": true,
+  "onnxRuntimeVersion": "1.29.0",
+  "torchVersion": "2.6.0"
+}

models/pl/kokoro-pl-fp32.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:66637ea517e8ea4fea6c473c18a6f3b4fe982a8e7eb3879bb5f8250c7dc5679a
+size 325563781

models/pl/kokoro-pl-fp32.validation.json

@@ -0,0 +1,23 @@
+{
+  "language": "pl-PL",
+  "model": "kokoro-pl-fp32.onnx",
+  "modelBytes": 325563781,
+  "modelSha256": "66637ea517e8ea4fea6c473c18a6f3b4fe982a8e7eb3879bb5f8250c7dc5679a",
+  "sourceCheckpoint": "kokoro_polish_converted.pth",
+  "sourceCheckpointSha256": "e3202dc4d1f6e65dddff8a8e8d2e091ce9a3a66cf989c6c581a81b2f1969af49",
+  "sourceRevision": "3744b57964eab7df6e8c48f0b84badb29e14df07",
+  "baseRevision": "f3ff3571791e39611d31c381e3a41a3af07b4987",
+  "opset": 17,
+  "sampleRateHz": 24000,
+  "voice": "pm_mateusz.bin",
+  "voiceSha256": "dc8f2919ede945e6962310b5204b912e64689d5698c571822bf75ad704870cb4",
+  "referencePhonemes": "z\u02c8a\u0292uwd\u0292 \u0261\u02c8\u025b\u0272\u0255l\u0254\u0303 j\u02c8a\u0292\u0272.",
+  "referenceInputTokens": 26,
+  "referenceOutputSamples": 49800,
+  "maxAbsoluteError": 0.07184462249279022,
+  "meanAbsoluteError": 0.0013195093488320708,
+  "waveformCorrelation": 0.9947539311111714,
+  "durationExactMatch": true,
+  "onnxRuntimeVersion": "1.29.0",
+  "torchVersion": "2.6.0"
+}

package-lock.json

@@ -0,0 +1,172 @@
+{
+  "name": "kokoro-kmp-model-validation",
+  "version": "1.0.0",
+  "lockfileVersion": 3,
+  "requires": true,
+  "packages": {
+    "": {
+      "name": "kokoro-kmp-model-validation",
+      "version": "1.0.0",
+      "devDependencies": {
+        "onnxruntime-web": "1.29.0"
+      }
+    },
+    "node_modules/@protobufjs/aspromise": {
+      "version": "1.1.2",
+      "resolved": "https://registry.npmjs.org/@protobufjs/aspromise/-/aspromise-1.1.2.tgz",
+      "integrity": "sha512-j+gKExEuLmKwvz3OgROXtrJ2UG2x8Ch2YZUxahh+s1F2HZ+wAceUNLkvy6zKCPVRkU++ZWQrdxsUeQXmcg4uoQ==",
+      "dev": true,
+      "license": "BSD-3-Clause"
+    },
+    "node_modules/@protobufjs/base64": {
+      "version": "1.1.2",
+      "resolved": "https://registry.npmjs.org/@protobufjs/base64/-/base64-1.1.2.tgz",
+      "integrity": "sha512-AZkcAA5vnN/v4PDqKyMR5lx7hZttPDgClv83E//FMNhR2TMcLUhfRUBHCmSl0oi9zMgDDqRUJkSxO3wm85+XLg==",
+      "dev": true,
+      "license": "BSD-3-Clause"
+    },
+    "node_modules/@protobufjs/codegen": {
+      "version": "2.0.5",
+      "resolved": "https://registry.npmjs.org/@protobufjs/codegen/-/codegen-2.0.5.tgz",
+      "integrity": "sha512-zgXFLzW3Ap33e6d0Wlj4MGIm6Ce8O89n/apUaGNB/jx+hw+ruWEp7EwGUshdLKVRCxZW12fp9r40E1mQrf/34g==",
+      "dev": true,
+      "license": "BSD-3-Clause"
+    },
+    "node_modules/@protobufjs/eventemitter": {
+      "version": "1.1.1",
+      "resolved": "https://registry.npmjs.org/@protobufjs/eventemitter/-/eventemitter-1.1.1.tgz",
+      "integrity": "sha512-vW1GmwMZNnL+gMRaovlh9yZX74kc+TTU3FObkkurpMaRtBfLP3ldjS9KQWlwZgraRE0+dheEEoAxdzcJQ8eXZg==",
+      "dev": true,
+      "license": "BSD-3-Clause"
+    },
+    "node_modules/@protobufjs/fetch": {
+      "version": "1.1.1",
+      "resolved": "https://registry.npmjs.org/@protobufjs/fetch/-/fetch-1.1.1.tgz",
+      "integrity": "sha512-GpptLrs57adMSuHi3VNj0mAF8dwh36LMaYF6XyJ6JMWlVsc+t42tm1HSEDmOs3A8fC9yyeisgLhsTVQokOZ0zw==",
+      "dev": true,
+      "license": "BSD-3-Clause",
+      "dependencies": {
+        "@protobufjs/aspromise": "^1.1.1"
+      }
+    },
+    "node_modules/@protobufjs/float": {
+      "version": "1.0.2",
+      "resolved": "https://registry.npmjs.org/@protobufjs/float/-/float-1.0.2.tgz",
+      "integrity": "sha512-Ddb+kVXlXst9d+R9PfTIxh1EdNkgoRe5tOX6t01f1lYWOvJnSPDBlG241QLzcyPdoNTsblLUdujGSE4RzrTZGQ==",
+      "dev": true,
+      "license": "BSD-3-Clause"
+    },
+    "node_modules/@protobufjs/path": {
+      "version": "1.1.2",
+      "resolved": "https://registry.npmjs.org/@protobufjs/path/-/path-1.1.2.tgz",
+      "integrity": "sha512-6JOcJ5Tm08dOHAbdR3GrvP+yUUfkjG5ePsHYczMFLq3ZmMkAD98cDgcT2iA1lJ9NVwFd4tH/iSSoe44YWkltEA==",
+      "dev": true,
+      "license": "BSD-3-Clause"
+    },
+    "node_modules/@protobufjs/pool": {
+      "version": "1.1.0",
+      "resolved": "https://registry.npmjs.org/@protobufjs/pool/-/pool-1.1.0.tgz",
+      "integrity": "sha512-0kELaGSIDBKvcgS4zkjz1PeddatrjYcmMWOlAuAPwAeccUrPHdUqo/J6LiymHHEiJT5NrF1UVwxY14f+fy4WQw==",
+      "dev": true,
+      "license": "BSD-3-Clause"
+    },
+    "node_modules/@protobufjs/utf8": {
+      "version": "1.1.2",
+      "resolved": "https://registry.npmjs.org/@protobufjs/utf8/-/utf8-1.1.2.tgz",
+      "integrity": "sha512-b1UQwcEZ4yCnMCD8DAL1VlbvBJE9/IX4FTIp7BG1xYpf29SLazLSrqUkj4w7Y5y7cCVP6E5tcqqcI0xemPkHug==",
+      "dev": true,
+      "license": "BSD-3-Clause"
+    },
+    "node_modules/@types/node": {
+      "version": "26.4.1",
+      "resolved": "https://registry.npmjs.org/@types/node/-/node-26.4.1.tgz",
+      "integrity": "sha512-k97ENvZWtvA6yqz5/FS6a7duDgOPEeOQOc2iKS/nY6mX6qJUKtLnWzQS+Xj6tXweyj6ZcTAK2Qecetnvi9nCLA==",
+      "dev": true,
+      "license": "MIT",
+      "dependencies": {
+        "undici-types": "~8.3.0"
+      }
+    },
+    "node_modules/flatbuffers": {
+      "version": "25.9.23",
+      "resolved": "https://registry.npmjs.org/flatbuffers/-/flatbuffers-25.9.23.tgz",
+      "integrity": "sha512-MI1qs7Lo4Syw0EOzUl0xjs2lsoeqFku44KpngfIduHBYvzm8h2+7K8YMQh1JtVVVrUvhLpNwqVi4DERegUJhPQ==",
+      "dev": true,
+      "license": "Apache-2.0"
+    },
+    "node_modules/guid-typescript": {
+      "version": "1.0.9",
+      "resolved": "https://registry.npmjs.org/guid-typescript/-/guid-typescript-1.0.9.tgz",
+      "integrity": "sha512-Y8T4vYhEfwJOTbouREvG+3XDsjr8E3kIr7uf+JZ0BYloFsttiHU0WfvANVsR7TxNUJa/WpCnw/Ino/p+DeBhBQ==",
+      "dev": true,
+      "license": "ISC"
+    },
+    "node_modules/long": {
+      "version": "5.3.2",
+      "resolved": "https://registry.npmjs.org/long/-/long-5.3.2.tgz",
+      "integrity": "sha512-mNAgZ1GmyNhD7AuqnTG3/VQ26o760+ZYBPKjPvugO8+nLbYfX6TVpJPseBvopbdY+qpZ/lKUnmEc1LeZYS3QAA==",
+      "dev": true,
+      "license": "Apache-2.0"
+    },
+    "node_modules/onnxruntime-common": {
+      "version": "1.29.0",
+      "resolved": "https://registry.npmjs.org/onnxruntime-common/-/onnxruntime-common-1.29.0.tgz",
+      "integrity": "sha512-/F63/e2VJoaVXGGNu6S5QH7jivBThGO95OzAVXXQ8hTta/b1QxI8udHa6cI3+3mAb5WWIIaMMwfZw01oivjJ1g==",
+      "dev": true,
+      "license": "MIT"
+    },
+    "node_modules/onnxruntime-web": {
+      "version": "1.29.0",
+      "resolved": "https://registry.npmjs.org/onnxruntime-web/-/onnxruntime-web-1.29.0.tgz",
+      "integrity": "sha512-LuQlpX6MFLJZu756erwUeb1mNfoJGbs1kzDwJGNlf5RvfYMdqhcY3vNpDPK40CUV2HoWTkIj+uS0o36GFHjeYw==",
+      "dev": true,
+      "license": "MIT",
+      "dependencies": {
+        "flatbuffers": "^25.1.24",
+        "guid-typescript": "^1.0.9",
+        "long": "^5.2.3",
+        "onnxruntime-common": "1.29.0",
+        "platform": "^1.3.6",
+        "protobufjs": "^7.2.4"
+      }
+    },
+    "node_modules/platform": {
+      "version": "1.3.6",
+      "resolved": "https://registry.npmjs.org/platform/-/platform-1.3.6.tgz",
+      "integrity": "sha512-fnWVljUchTro6RiCFvCXBbNhJc2NijN7oIQxbwsyL0buWJPG85v81ehlHI9fXrJsMNgTofEoWIQeClKpgxFLrg==",
+      "dev": true,
+      "license": "MIT"
+    },
+    "node_modules/protobufjs": {
+      "version": "7.6.6",
+      "resolved": "https://registry.npmjs.org/protobufjs/-/protobufjs-7.6.6.tgz",
+      "integrity": "sha512-dYDWdjSl5RNb7SgPxGQcRU+GtvP7s2fpkrY0r432PcOIaZ0/rBcxEZnQN67iJhFuQiVw754JDoPruPCNdGsbjg==",
+      "dev": true,
+      "hasInstallScript": true,
+      "license": "BSD-3-Clause",
+      "dependencies": {
+        "@protobufjs/aspromise": "^1.1.2",
+        "@protobufjs/base64": "^1.1.2",
+        "@protobufjs/codegen": "^2.0.5",
+        "@protobufjs/eventemitter": "^1.1.1",
+        "@protobufjs/fetch": "^1.1.1",
+        "@protobufjs/float": "^1.0.2",
+        "@protobufjs/path": "^1.1.2",
+        "@protobufjs/pool": "^1.1.0",
+        "@protobufjs/utf8": "^1.1.1",
+        "@types/node": ">=13.7.0",
+        "long": "^5.3.2"
+      },
+      "engines": {
+        "node": ">=12.0.0"
+      }
+    },
+    "node_modules/undici-types": {
+      "version": "8.3.0",
+      "resolved": "https://registry.npmjs.org/undici-types/-/undici-types-8.3.0.tgz",
+      "integrity": "sha512-j375ScV60dom+YkPFIfTLcOiPxkN/buHz5GobjLhixFuANaNs3C9l4GmrWqejgXWJ7BbJcFYpTEUkS1Ge8bpZQ==",
+      "dev": true,
+      "license": "MIT"
+    }
+  }
+}

package.json

@@ -0,0 +1,13 @@
+{
+  "name": "kokoro-kmp-model-validation",
+  "version": "1.0.0",
+  "private": true,
+  "scripts": {
+    "validate:web": "node scripts/validate_web.cjs pl && node scripts/validate_web.cjs de",
+    "validate:web:pl": "node scripts/validate_web.cjs pl",
+    "validate:web:de": "node scripts/validate_web.cjs de"
+  },
+  "devDependencies": {
+    "onnxruntime-web": "1.29.0"
+  }
+}

phonemizers/de/phonemizer_de.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4888dc7e54dc66098551555096562063364091fc246da0d057b629587bedaa0b
+size 7094120

phonemizers/pl/phonemizer_pl.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ec85f4dc2c4ac7a72ff88b98b0664a4ed887bd15c0d5add2eb1d6a6ee05b73f2
+size 7094120

requirements-export.txt

@@ -0,0 +1,7 @@
+numpy==2.5.2
+huggingface-hub==0.36.2
+jsonschema==4.26.0
+onnx==1.19.0
+onnxruntime==1.29.0
+soundfile==0.13.1
+torch==2.6.0

runtime/config.json

@@ -0,0 +1,150 @@
+{
+  "istftnet": {
+    "upsample_kernel_sizes": [20, 12],
+    "upsample_rates": [10, 6],
+    "gen_istft_hop_size": 5,
+    "gen_istft_n_fft": 20,
+    "resblock_dilation_sizes": [
+      [1, 3, 5],
+      [1, 3, 5],
+      [1, 3, 5]
+    ],
+    "resblock_kernel_sizes": [3, 7, 11],
+    "upsample_initial_channel": 512
+  },
+  "dim_in": 64,
+  "dropout": 0.2,
+  "hidden_dim": 512,
+  "max_conv_dim": 512,
+  "max_dur": 50,
+  "multispeaker": true,
+  "n_layer": 3,
+  "n_mels": 80,
+  "n_token": 178,
+  "style_dim": 128,
+  "text_encoder_kernel_size": 5,
+  "plbert": {
+    "hidden_size": 768,
+    "num_attention_heads": 12,
+    "intermediate_size": 2048,
+    "max_position_embeddings": 512,
+    "num_hidden_layers": 12,
+    "dropout": 0.1
+  },
+  "vocab": {
+    ";": 1,
+    ":": 2,
+    ",": 3,
+    ".": 4,
+    "!": 5,
+    "?": 6,
+    "—": 9,
+    "…": 10,
+    "\"": 11,
+    "(": 12,
+    ")": 13,
+    "“": 14,
+    "”": 15,
+    " ": 16,
+    "\u0303": 17,
+    "ʣ": 18,
+    "ʥ": 19,
+    "ʦ": 20,
+    "ʨ": 21,
+    "ᵝ": 22,
+    "\uAB67": 23,
+    "A": 24,
+    "I": 25,
+    "O": 31,
+    "Q": 33,
+    "S": 35,
+    "T": 36,
+    "W": 39,
+    "Y": 41,
+    "ᵊ": 42,
+    "a": 43,
+    "b": 44,
+    "c": 45,
+    "d": 46,
+    "e": 47,
+    "f": 48,
+    "h": 50,
+    "i": 51,
+    "j": 52,
+    "k": 53,
+    "l": 54,
+    "m": 55,
+    "n": 56,
+    "o": 57,
+    "p": 58,
+    "q": 59,
+    "r": 60,
+    "s": 61,
+    "t": 62,
+    "u": 63,
+    "v": 64,
+    "w": 65,
+    "x": 66,
+    "y": 67,
+    "z": 68,
+    "ɑ": 69,
+    "ɐ": 70,
+    "ɒ": 71,
+    "æ": 72,
+    "β": 75,
+    "ɔ": 76,
+    "ɕ": 77,
+    "ç": 78,
+    "ɖ": 80,
+    "ð": 81,
+    "ʤ": 82,
+    "ə": 83,
+    "ɚ": 85,
+    "ɛ": 86,
+    "ɜ": 87,
+    "ɟ": 90,
+    "ɡ": 92,
+    "ɥ": 99,
+    "ɨ": 101,
+    "ɪ": 102,
+    "ʝ": 103,
+    "ɯ": 110,
+    "ɰ": 111,
+    "ŋ": 112,
+    "ɳ": 113,
+    "ɲ": 114,
+    "ɴ": 115,
+    "ø": 116,
+    "ɸ": 118,
+    "θ": 119,
+    "œ": 120,
+    "ɹ": 123,
+    "ɾ": 125,
+    "ɻ": 126,
+    "ʁ": 128,
+    "ɽ": 129,
+    "ʂ": 130,
+    "ʃ": 131,
+    "ʈ": 132,
+    "ʧ": 133,
+    "ʊ": 135,
+    "ʋ": 136,
+    "ʌ": 138,
+    "ɣ": 139,
+    "ɤ": 140,
+    "χ": 142,
+    "ʎ": 143,
+    "ʒ": 147,
+    "ʔ": 148,
+    "ˈ": 156,
+    "ˌ": 157,
+    "ː": 158,
+    "ʰ": 162,
+    "ʲ": 164,
+    "↓": 169,
+    "→": 171,
+    "↗": 172,
+    "↘": 173,
+    "ᵻ": 177
+  }
+}

samples/de-anna.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ab75032aeb5be87b1844fb7008c58c0bda40ca52f907ed356466e11abb4adcab
+size 180044

schema/kokoro-catalog.schema.json

@@ -0,0 +1,157 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "title": "Kokoro KMP model catalog",
+  "type": "object",
+  "additionalProperties": false,
+  "required": [
+    "schemaVersion",
+    "catalogVersion",
+    "distribution",
+    "license",
+    "modelFamily",
+    "audio",
+    "runtime",
+    "languages"
+  ],
+  "properties": {
+    "$schema": { "type": "string" },
+    "schemaVersion": { "const": 1 },
+    "catalogVersion": { "type": "string", "pattern": "^[0-9]+\\.[0-9]+\\.[0-9]+$" },
+    "distribution": {
+      "type": "object",
+      "additionalProperties": false,
+      "required": ["provider", "artifactResolution", "recommendedRevision"],
+      "properties": {
+        "provider": { "const": "huggingface" },
+        "artifactResolution": { "const": "relative-to-catalog" },
+        "recommendedRevision": { "type": "string", "minLength": 1 }
+      }
+    },
+    "license": { "const": "Apache-2.0" },
+    "modelFamily": { "const": "kokoro-82m" },
+    "audio": {
+      "type": "object",
+      "additionalProperties": false,
+      "required": ["sampleRateHz", "channels", "sampleType"],
+      "properties": {
+        "sampleRateHz": { "const": 24000 },
+        "channels": { "const": 1 },
+        "sampleType": { "const": "float32" }
+      }
+    },
+    "runtime": {
+      "type": "object",
+      "additionalProperties": true,
+      "required": [
+        "engine",
+        "opset",
+        "batchSize",
+        "maxPhonemeCodePoints",
+        "inputs",
+        "outputs",
+        "tokenizer",
+        "voiceStyle"
+      ],
+      "properties": {
+        "engine": { "const": "onnxruntime" },
+        "opset": { "type": "integer", "minimum": 1 },
+        "batchSize": { "const": 1 },
+        "maxPhonemeCodePoints": { "type": "integer", "minimum": 1 },
+        "tokenizer": {
+          "type": "object",
+          "required": ["bosTokenId", "eosTokenId", "vocabularyField", "artifact"],
+          "properties": {
+            "artifact": { "$ref": "#/$defs/artifact" }
+          }
+        },
+        "voiceStyle": {
+          "type": "object",
+          "required": ["format", "shape", "rowSelection"]
+        }
+      }
+    },
+    "languages": {
+      "type": "array",
+      "minItems": 1,
+      "items": { "$ref": "#/$defs/language" }
+    }
+  },
+  "$defs": {
+    "artifact": {
+      "type": "object",
+      "additionalProperties": false,
+      "required": ["path", "mediaType", "sizeBytes", "sha256"],
+      "properties": {
+        "path": {
+          "type": "string",
+          "minLength": 1,
+          "not": { "pattern": "(^/|(^|/)\\.\\.(/|$))" }
+        },
+        "mediaType": { "type": "string", "minLength": 1 },
+        "sizeBytes": { "type": "integer", "minimum": 0 },
+        "sha256": { "type": "string", "pattern": "^[a-f0-9]{64}$" }
+      }
+    },
+    "voice": {
+      "type": "object",
+      "additionalProperties": false,
+      "required": ["id", "displayName", "gender", "artifact"],
+      "properties": {
+        "id": { "type": "string", "pattern": "^[a-z0-9_-]+$" },
+        "displayName": { "type": "string", "minLength": 1 },
+        "gender": { "enum": ["female", "male", "neutral", "unknown"] },
+        "artifact": { "$ref": "#/$defs/artifact" }
+      }
+    },
+    "language": {
+      "type": "object",
+      "additionalProperties": false,
+      "required": [
+        "id",
+        "locale",
+        "displayName",
+        "nativeName",
+        "model",
+        "phonemizer",
+        "voices",
+        "previewAudioPath"
+      ],
+      "properties": {
+        "id": { "type": "string", "pattern": "^[a-z]{2,3}$" },
+        "locale": { "type": "string", "minLength": 2 },
+        "displayName": { "type": "string", "minLength": 1 },
+        "nativeName": { "type": "string", "minLength": 1 },
+        "model": {
+          "type": "object",
+          "additionalProperties": false,
+          "required": ["id", "format", "precision", "artifact", "validation"],
+          "properties": {
+            "id": { "type": "string", "minLength": 1 },
+            "format": { "const": "onnx" },
+            "precision": { "enum": ["fp32", "fp16", "int8"] },
+            "artifact": { "$ref": "#/$defs/artifact" },
+            "validation": { "$ref": "#/$defs/artifact" }
+          }
+        },
+        "phonemizer": {
+          "type": "object",
+          "additionalProperties": false,
+          "required": ["engine", "language", "license", "sourceRevision", "artifact"],
+          "properties": {
+            "engine": { "const": "phonemis-protophone" },
+            "language": { "type": "string", "minLength": 2 },
+            "license": { "const": "MIT" },
+            "sourceRevision": { "type": "string", "pattern": "^[a-f0-9]{40}$" },
+            "artifact": { "$ref": "#/$defs/artifact" }
+          }
+        },
+        "voices": {
+          "type": "array",
+          "minItems": 1,
+          "items": { "$ref": "#/$defs/voice" }
+        },
+        "previewAudioPath": { "type": "string", "minLength": 1 }
+      }
+    }
+  }
+}

scripts/build_catalog.py

@@ -0,0 +1,50 @@
+#!/usr/bin/env python3
+"""Hydrate catalog.template.json with immutable artifact sizes and checksums."""
+
+from __future__ import annotations
+
+import hashlib
+import json
+from pathlib import Path
+from typing import Any
+
+
+ROOT = Path(__file__).resolve().parents[1]
+
+
+def sha256(path: Path) -> str:
+    digest = hashlib.sha256()
+    with path.open("rb") as source:
+        for block in iter(lambda: source.read(1024 * 1024), b""):
+            digest.update(block)
+    return digest.hexdigest()
+
+
+def hydrate(value: Any) -> Any:
+    if isinstance(value, list):
+        return [hydrate(item) for item in value]
+    if not isinstance(value, dict):
+        return value
+    if "$artifact" in value:
+        if len(value) != 1 or not isinstance(value["$artifact"], dict):
+            raise ValueError("$artifact must be the only key and contain an object")
+        artifact = dict(value["$artifact"])
+        relative_path = Path(artifact["path"])
+        if relative_path.is_absolute() or ".." in relative_path.parts:
+            raise ValueError(f"Artifact path must stay inside the repository: {relative_path}")
+        path = ROOT / relative_path
+        if not path.is_file():
+            raise FileNotFoundError(path)
+        artifact["sizeBytes"] = path.stat().st_size
+        artifact["sha256"] = sha256(path)
+        return artifact
+    return {key: hydrate(item) for key, item in value.items()}
+
+
+template = json.loads((ROOT / "catalog.template.json").read_text(encoding="utf-8"))
+catalog = hydrate(template)
+(ROOT / "catalog.json").write_text(
+    json.dumps(catalog, ensure_ascii=False, indent=2) + "\n",
+    encoding="utf-8",
+)
+print(ROOT / "catalog.json")

scripts/export_onnx.py

@@ -0,0 +1,201 @@
+#!/usr/bin/env python3
+"""Export a Kokoro checkpoint to a self-contained ONNX model and verify it."""
+
+from __future__ import annotations
+
+import argparse
+import hashlib
+import json
+from pathlib import Path
+
+import numpy as np
+import onnx
+import onnxruntime as ort
+import torch
+
+from kokoro import KModel
+from kokoro.model import KModelForONNX
+
+
+def sha256(path: Path) -> str:
+    digest = hashlib.sha256()
+    with path.open("rb") as source:
+        for block in iter(lambda: source.read(1024 * 1024), b""):
+            digest.update(block)
+    return digest.hexdigest()
+
+
+def make_reference_inputs(
+    model: KModel,
+    voice: Path,
+    phonemes: str,
+) -> tuple[torch.Tensor, torch.Tensor, torch.Tensor]:
+    missing = sorted({phoneme for phoneme in phonemes if model.vocab.get(phoneme) is None})
+    if missing:
+        raise RuntimeError(f"Reference phonemes missing from the Kokoro vocabulary: {missing}")
+    if not 1 <= len(phonemes) <= 510:
+        raise RuntimeError(f"Reference phoneme length must be in 1..510, got {len(phonemes)}")
+
+    voice_pack = np.fromfile(voice, dtype=np.float32)
+    if voice_pack.size != 510 * 256:
+        raise RuntimeError(
+            f"Expected a raw 510x256 float32 voice pack, got {voice_pack.size} values"
+        )
+    voice_pack = voice_pack.reshape(510, 256)
+
+    token_ids = [int(model.vocab[phoneme]) for phoneme in phonemes]
+    input_ids = torch.tensor([[0, *token_ids, 0]], dtype=torch.long)
+    style = torch.from_numpy(voice_pack[len(phonemes) - 1]).unsqueeze(0)
+    speed = torch.tensor([1.0], dtype=torch.float32)
+    return input_ids, style, speed
+
+
+def add_metadata(
+    output: Path,
+    *,
+    language: str,
+    checkpoint: Path,
+    source_revision: str,
+    base_revision: str,
+) -> None:
+    model = onnx.load(output)
+    onnx.helper.set_model_props(
+        model,
+        {
+            "architecture": "Kokoro-82M",
+            "language": language,
+            "license": "Apache-2.0",
+            "sample_rate_hz": "24000",
+            "source_repository": "software-mansion/react-native-executorch-kokoro",
+            "source_revision": source_revision,
+            "source_checkpoint": checkpoint.name,
+            "source_checkpoint_sha256": sha256(checkpoint),
+            "base_repository": "hexgrad/Kokoro-82M",
+            "base_revision": base_revision,
+        },
+    )
+    onnx.save_model(model, output)
+
+
+def export(args: argparse.Namespace) -> None:
+    checkpoint = args.checkpoint.resolve()
+    config = args.config.resolve()
+    voice = args.voice.resolve()
+    output = args.output.resolve()
+    output.parent.mkdir(parents=True, exist_ok=True)
+
+    torch.manual_seed(0)
+    model = KModel(
+        repo_id="hexgrad/Kokoro-82M",
+        config=str(config),
+        model=str(checkpoint),
+        disable_complex=True,
+    ).cpu().eval()
+    export_model = KModelForONNX(model).cpu().eval()
+    input_ids, style, speed = make_reference_inputs(
+        model,
+        voice,
+        args.reference_phonemes,
+    )
+
+    torch.onnx.export(
+        export_model,
+        args=(input_ids, style, speed),
+        f=str(output),
+        input_names=["input_ids", "style", "speed"],
+        output_names=["waveform", "duration"],
+        opset_version=17,
+        dynamic_axes={
+            "input_ids": {1: "sequence_length"},
+            "waveform": {0: "num_samples"},
+            "duration": {0: "sequence_length"},
+        },
+        do_constant_folding=True,
+        dynamo=False,
+    )
+
+    add_metadata(
+        output,
+        language=args.language,
+        checkpoint=checkpoint,
+        source_revision=args.source_revision,
+        base_revision=args.base_revision,
+    )
+    onnx.checker.check_model(str(output), full_check=True)
+
+    options = ort.SessionOptions()
+    options.graph_optimization_level = ort.GraphOptimizationLevel.ORT_ENABLE_ALL
+    session = ort.InferenceSession(str(output), sess_options=options, providers=["CPUExecutionProvider"])
+
+    # Kokoro deliberately injects a small amount of random excitation noise. As a
+    # result, PyTorch and ONNX waveforms should be highly correlated, but are not
+    # expected to be bit-identical or pass a point-wise allclose assertion.
+    torch.manual_seed(0)
+    with torch.inference_mode():
+        torch_waveform, torch_duration = export_model(input_ids, style, speed)
+    ort_waveform, ort_duration = session.run(
+        None,
+        {
+            "input_ids": input_ids.numpy(),
+            "style": style.numpy(),
+            "speed": speed.numpy(),
+        },
+    )
+
+    expected_waveform = torch_waveform.detach().cpu().numpy()
+    expected_duration = torch_duration.detach().cpu().numpy()
+    np.testing.assert_array_equal(ort_duration, expected_duration)
+    if not np.isfinite(ort_waveform).all():
+        raise RuntimeError("ONNX Runtime produced non-finite waveform samples")
+
+    delta = np.abs(ort_waveform - expected_waveform)
+    correlation = float(np.corrcoef(ort_waveform, expected_waveform)[0, 1])
+    mean_absolute_error = float(delta.mean())
+    if correlation < 0.99 or mean_absolute_error > 0.005:
+        raise RuntimeError(
+            "ONNX/PyTorch validation failed: "
+            f"correlation={correlation:.6f}, meanAbsoluteError={mean_absolute_error:.6f}"
+        )
+    validation = {
+        "language": args.language,
+        "model": output.name,
+        "modelBytes": output.stat().st_size,
+        "modelSha256": sha256(output),
+        "sourceCheckpoint": checkpoint.name,
+        "sourceCheckpointSha256": sha256(checkpoint),
+        "sourceRevision": args.source_revision,
+        "baseRevision": args.base_revision,
+        "opset": 17,
+        "sampleRateHz": 24000,
+        "voice": voice.name,
+        "voiceSha256": sha256(voice),
+        "referencePhonemes": args.reference_phonemes,
+        "referenceInputTokens": int(input_ids.shape[1]),
+        "referenceOutputSamples": int(ort_waveform.size),
+        "maxAbsoluteError": float(delta.max(initial=0.0)),
+        "meanAbsoluteError": mean_absolute_error,
+        "waveformCorrelation": correlation,
+        "durationExactMatch": True,
+        "onnxRuntimeVersion": ort.__version__,
+        "torchVersion": torch.__version__,
+    }
+    validation_path = output.with_suffix(".validation.json")
+    validation_path.write_text(json.dumps(validation, indent=2) + "\n", encoding="utf-8")
+    print(json.dumps(validation, indent=2))
+
+
+def parse_args() -> argparse.Namespace:
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--checkpoint", type=Path, required=True)
+    parser.add_argument("--config", type=Path, required=True)
+    parser.add_argument("--voice", type=Path, required=True)
+    parser.add_argument("--output", type=Path, required=True)
+    parser.add_argument("--language", required=True)
+    parser.add_argument("--reference-phonemes", required=True)
+    parser.add_argument("--source-revision", required=True)
+    parser.add_argument("--base-revision", required=True)
+    return parser.parse_args()
+
+
+if __name__ == "__main__":
+    export(parse_args())

scripts/publish_hf.py

@@ -0,0 +1,88 @@
+#!/usr/bin/env python3
+"""Validate and publish this folder as a public Hugging Face model repository."""
+
+from __future__ import annotations
+
+import argparse
+import subprocess
+import sys
+from pathlib import Path
+
+from huggingface_hub import HfApi
+
+
+ROOT = Path(__file__).resolve().parents[1]
+EXPECTED_REPOSITORY_NAME = "kokoro-kmp-models"
+
+
+def run_validation(skip_web: bool) -> None:
+    subprocess.run([sys.executable, ROOT / "scripts/build_catalog.py"], check=True, cwd=ROOT)
+    subprocess.run([sys.executable, ROOT / "scripts/validate_package.py"], check=True, cwd=ROOT)
+    if not skip_web:
+        subprocess.run(["npm", "run", "validate:web"], check=True, cwd=ROOT)
+
+
+def parse_args() -> argparse.Namespace:
+    parser = argparse.ArgumentParser()
+    parser.add_argument("repo_id", help="Hugging Face repo, e.g. username/kokoro-kmp-models")
+    parser.add_argument("--tag", default="v1.0.0")
+    parser.add_argument("--skip-web-validation", action="store_true")
+    return parser.parse_args()
+
+
+def main() -> None:
+    args = parse_args()
+    parts = args.repo_id.split("/")
+    if len(parts) != 2 or parts[1] != EXPECTED_REPOSITORY_NAME:
+        raise SystemExit(
+            f"repo_id must be <namespace>/{EXPECTED_REPOSITORY_NAME}; got {args.repo_id!r}"
+        )
+
+    run_validation(args.skip_web_validation)
+
+    api = HfApi()
+    identity = api.whoami()
+    writable_namespaces = {identity["name"]}
+    writable_namespaces.update(
+        org["name"]
+        for org in identity.get("orgs", [])
+        if org.get("roleInOrg") in {"admin", "write"}
+    )
+    if parts[0] not in writable_namespaces:
+        raise SystemExit(f"The authenticated account cannot publish to namespace {parts[0]!r}")
+
+    api.create_repo(args.repo_id, repo_type="model", private=False, exist_ok=True)
+    commit = api.upload_folder(
+        repo_id=args.repo_id,
+        repo_type="model",
+        folder_path=ROOT,
+        commit_message=f"Publish Kokoro PL/DE model catalog {args.tag}",
+        ignore_patterns=[
+            ".git/**",
+            "node_modules/**",
+            "**/__pycache__/**",
+            "**/*.pyc",
+            ":memory:.ses",
+        ],
+    )
+
+    tags = {tag.ref: tag.target_commit for tag in api.list_repo_refs(args.repo_id).tags}
+    tag_ref = f"refs/tags/{args.tag}"
+    existing_target = tags.get(tag_ref)
+    if existing_target and existing_target != commit.oid:
+        raise SystemExit(
+            f"Refusing to move immutable tag {args.tag}: it already points to {existing_target}"
+        )
+    if not existing_target:
+        api.create_tag(
+            args.repo_id,
+            repo_type="model",
+            tag=args.tag,
+            revision=commit.oid,
+            tag_message=f"Kokoro PL/DE catalog {args.tag}",
+        )
+    print(f"https://huggingface.co/{args.repo_id}/tree/{args.tag}")
+
+
+if __name__ == "__main__":
+    main()

scripts/validate_package.py

@@ -0,0 +1,159 @@
+#!/usr/bin/env python3
+"""Validate every catalog artifact and exercise all ONNX language packs."""
+
+from __future__ import annotations
+
+import hashlib
+import json
+from pathlib import Path
+from typing import Any, Iterator
+
+import numpy as np
+import onnxruntime as ort
+import soundfile as sf
+from jsonschema import Draft202012Validator
+
+
+ROOT = Path(__file__).resolve().parents[1]
+
+
+def sha256(path: Path) -> str:
+    digest = hashlib.sha256()
+    with path.open("rb") as source:
+        for block in iter(lambda: source.read(1024 * 1024), b""):
+            digest.update(block)
+    return digest.hexdigest()
+
+
+def artifacts(value: Any) -> Iterator[dict[str, Any]]:
+    if isinstance(value, list):
+        for item in value:
+            yield from artifacts(item)
+    elif isinstance(value, dict):
+        if {"path", "sizeBytes", "sha256"} <= value.keys():
+            yield value
+        for item in value.values():
+            yield from artifacts(item)
+
+
+catalog = json.loads((ROOT / "catalog.json").read_text(encoding="utf-8"))
+schema = json.loads(
+    (ROOT / "schema/kokoro-catalog.schema.json").read_text(encoding="utf-8")
+)
+Draft202012Validator.check_schema(schema)
+Draft202012Validator(schema).validate(catalog)
+for artifact in artifacts(catalog):
+    path = ROOT / artifact["path"]
+    if not path.is_file():
+        raise FileNotFoundError(path)
+    if path.stat().st_size != artifact["sizeBytes"]:
+        raise RuntimeError(f"Size mismatch: {path}")
+    if sha256(path) != artifact["sha256"]:
+        raise RuntimeError(f"SHA-256 mismatch: {path}")
+
+tokenizer = catalog["runtime"]["tokenizer"]
+config_artifact = tokenizer["artifact"]
+config = json.loads((ROOT / config_artifact["path"]).read_text(encoding="utf-8"))
+vocab = config[tokenizer["vocabularyField"]]
+languages = {language["id"]: language for language in catalog["languages"]}
+cases = json.loads((ROOT / "validation/cases.json").read_text(encoding="utf-8"))
+sessions: dict[str, ort.InferenceSession] = {}
+results = []
+
+for case in cases:
+    language = languages[case["language"]]
+    voices = {voice["id"]: voice for voice in language["voices"]}
+    voice = voices[case["voice"]]
+    phonemes = case["phonemes"]
+    missing = sorted({phoneme for phoneme in phonemes if phoneme not in vocab})
+    if missing:
+        raise RuntimeError(f"{case['id']}: phonemes missing from vocabulary: {missing}")
+    if not 1 <= len(phonemes) <= catalog["runtime"]["maxPhonemeCodePoints"]:
+        raise RuntimeError(f"{case['id']}: invalid phoneme length {len(phonemes)}")
+
+    token_ids = [vocab[phoneme] for phoneme in phonemes]
+    input_ids = np.asarray([[0, *token_ids, 0]], dtype=np.int64)
+    voice_pack = np.fromfile(ROOT / voice["artifact"]["path"], dtype="<f4").reshape(510, 256)
+    style = voice_pack[len(phonemes) - 1].reshape(1, 256)
+    speed = np.asarray([1.0], dtype=np.float32)
+
+    model_path = language["model"]["artifact"]["path"]
+    if model_path not in sessions:
+        sessions[model_path] = ort.InferenceSession(
+            str(ROOT / model_path),
+            providers=["CPUExecutionProvider"],
+        )
+    waveform, duration = sessions[model_path].run(
+        None,
+        {"input_ids": input_ids, "style": style, "speed": speed},
+    )
+    if not np.isfinite(waveform).all():
+        raise RuntimeError(f"{case['id']}: non-finite waveform")
+    if duration.shape != (input_ids.shape[1],):
+        raise RuntimeError(f"{case['id']}: unexpected duration shape {duration.shape}")
+    expected_samples = int(duration.sum()) * catalog["runtime"]["durationToSamplesFactor"]
+    if waveform.shape != (expected_samples,):
+        raise RuntimeError(
+            f"{case['id']}: expected {expected_samples} samples, got {waveform.shape}"
+        )
+
+    if not case["id"].endswith("-dynamic"):
+        preview_path = ROOT / language["previewAudioPath"]
+        preview_path.parent.mkdir(parents=True, exist_ok=True)
+        sf.write(preview_path, waveform, catalog["audio"]["sampleRateHz"], subtype="PCM_16")
+
+    results.append(
+        {
+            "id": case["id"],
+            "phonemeCodePoints": len(phonemes),
+            "inputTokens": int(input_ids.shape[1]),
+            "outputSamples": int(waveform.size),
+            "durationSeconds": waveform.size / catalog["audio"]["sampleRateHz"],
+            "peakAbsoluteAmplitude": float(np.abs(waveform).max(initial=0.0)),
+        }
+    )
+
+boundary_results = []
+for language in catalog["languages"]:
+    voice = language["voices"][0]
+    voice_pack = np.fromfile(ROOT / voice["artifact"]["path"], dtype="<f4").reshape(510, 256)
+    model_path = language["model"]["artifact"]["path"]
+    session = sessions[model_path]
+    for phoneme_count in (1, 128, catalog["runtime"]["maxPhonemeCodePoints"]):
+        token_ids = [vocab["a"]] * phoneme_count
+        input_ids = np.asarray([[0, *token_ids, 0]], dtype=np.int64)
+        waveform, duration = session.run(
+            None,
+            {
+                "input_ids": input_ids,
+                "style": voice_pack[phoneme_count - 1].reshape(1, 256),
+                "speed": np.asarray([1.0], dtype=np.float32),
+            },
+        )
+        expected_samples = int(duration.sum()) * catalog["runtime"]["durationToSamplesFactor"]
+        if waveform.shape != (expected_samples,) or not np.isfinite(waveform).all():
+            raise RuntimeError(
+                f"{language['id']}: failed boundary case with {phoneme_count} phonemes"
+            )
+        boundary_results.append(
+            {
+                "language": language["id"],
+                "phonemeCodePoints": phoneme_count,
+                "inputTokens": int(input_ids.shape[1]),
+                "outputSamples": int(waveform.size),
+            }
+        )
+
+report = {
+    "catalogVersion": catalog["catalogVersion"],
+    "catalogSchemaValidated": True,
+    "onnxRuntimeVersion": ort.__version__,
+    "artifactsVerified": sum(1 for _ in artifacts(catalog)),
+    "cases": results,
+    "boundaryCases": boundary_results,
+}
+(ROOT / "validation/package-validation.json").write_text(
+    json.dumps(report, ensure_ascii=False, indent=2) + "\n",
+    encoding="utf-8",
+)
+print(json.dumps(report, ensure_ascii=False, indent=2))

scripts/validate_web.cjs

@@ -0,0 +1,107 @@
+#!/usr/bin/env node
+
+const fs = require('node:fs');
+const path = require('node:path');
+const ort = require('onnxruntime-web');
+
+const root = path.resolve(__dirname, '..');
+const languageId = process.argv[2];
+if (!languageId) throw new Error('usage: node scripts/validate_web.cjs <pl|de>');
+
+ort.env.wasm.numThreads = 1;
+ort.env.wasm.proxy = false;
+
+async function main() {
+  const catalog = JSON.parse(fs.readFileSync(path.join(root, 'catalog.json'), 'utf8'));
+  const cases = JSON.parse(fs.readFileSync(path.join(root, 'validation/cases.json'), 'utf8'));
+  const config = JSON.parse(fs.readFileSync(path.join(root, 'runtime/config.json'), 'utf8'));
+  const language = catalog.languages.find((item) => item.id === languageId);
+  const testCase = cases.find(
+    (item) => item.language === languageId && item.id.endsWith('-dynamic'),
+  );
+  if (!language || !testCase) throw new Error(`unknown language: ${languageId}`);
+  const voice = language.voices.find((item) => item.id === testCase.voice);
+
+  const phonemes = Array.from(testCase.phonemes);
+  const ids = [
+    0,
+    ...phonemes.map((phoneme) => {
+      const id = config.vocab[phoneme];
+      if (id === undefined) throw new Error(`missing phoneme: ${phoneme}`);
+      return id;
+    }),
+    0,
+  ];
+
+  const voiceBytes = fs.readFileSync(path.join(root, voice.artifact.path));
+  function styleAt(row) {
+    const styleOffset = row * 256 * 4;
+    const style = new Float32Array(256);
+    for (let index = 0; index < style.length; index += 1) {
+      style[index] = voiceBytes.readFloatLE(styleOffset + index * 4);
+    }
+    return style;
+  }
+
+  const modelBytes = fs.readFileSync(path.join(root, language.model.artifact.path));
+  const session = await ort.InferenceSession.create(modelBytes, {
+    executionProviders: ['wasm'],
+    graphOptimizationLevel: 'all',
+  });
+
+  async function validate(caseId, caseIds, style) {
+    const output = await session.run({
+      input_ids: new ort.Tensor(
+        'int64',
+        BigInt64Array.from(caseIds.map(BigInt)),
+        [1, caseIds.length],
+      ),
+      style: new ort.Tensor('float32', style, [1, 256]),
+      speed: new ort.Tensor('float32', Float32Array.of(1), [1]),
+    });
+    const waveform = output.waveform.data;
+    const durationSum = Array.from(output.duration.data).reduce(
+      (sum, item) => sum + Number(item),
+      0,
+    );
+    const expectedSamples = durationSum * catalog.runtime.durationToSamplesFactor;
+    if (waveform.length !== expectedSamples) {
+      throw new Error(`waveform/duration mismatch: ${waveform.length} vs ${expectedSamples}`);
+    }
+    let peak = 0;
+    for (const sample of waveform) {
+      if (!Number.isFinite(sample)) throw new Error('non-finite waveform');
+      peak = Math.max(peak, Math.abs(sample));
+    }
+    return {
+      id: caseId,
+      inputTokens: caseIds.length,
+      outputSamples: waveform.length,
+      peakAbsoluteAmplitude: peak,
+    };
+  }
+
+  const maxPhonemes = catalog.runtime.maxPhonemeCodePoints;
+  const maxIds = [0, ...Array(maxPhonemes).fill(config.vocab.a), 0];
+  const result = {
+    language: languageId,
+    onnxRuntimeWebVersion: JSON.parse(
+      fs.readFileSync(path.join(root, 'node_modules/onnxruntime-web/package.json'), 'utf8'),
+    ).version,
+    executionProvider: 'wasm',
+    cases: [
+      await validate(testCase.id, ids, styleAt(phonemes.length - 1)),
+      await validate('max-phoneme-boundary', maxIds, styleAt(maxPhonemes - 1)),
+    ],
+  };
+  fs.writeFileSync(
+    path.join(root, 'validation', `web-${languageId}.json`),
+    `${JSON.stringify(result, null, 2)}\n`,
+  );
+  console.log(JSON.stringify(result));
+}
+
+main().catch((error) => {
+  console.error(error);
+  process.exitCode = 1;
+});

third-party/kokoro/LICENSE

@@ -0,0 +1,201 @@
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

third-party/phonemis/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 IgorSwat
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

validation/cases.json

@@ -0,0 +1,30 @@
+[
+  {
+    "id": "pl-mateusz",
+    "language": "pl",
+    "voice": "mateusz",
+    "text": "Zażółć gęślą jaźń.",
+    "phonemes": "zˈaʒuwtɕ ɡˈɛɲɕlɔ̃ jˈaʒɲ."
+  },
+  {
+    "id": "pl-dynamic",
+    "language": "pl",
+    "voice": "mateusz",
+    "text": "W Szczebrzeszynie chrząszcz brzmi w trzcinie.",
+    "phonemes": "f ʃʧˌɛbʒɛʃˈɨɲʲɛ xʃˈɔ̃ʃʧ bʒmˈi f tʃtɕˈiɲʲɛ."
+  },
+  {
+    "id": "de-anna",
+    "language": "de",
+    "voice": "anna",
+    "text": "Falsches Üben von Xylophonmusik quält jeden größeren Zwerg.",
+    "phonemes": "fˈalʃəs ˈyːbən fˈɔn ksˈyːloːfˌɔnmuːzˌiːk kvˈɛlt jˈeːdən ɡɾˈøːsərən ʦvˈɛɾk."
+  },
+  {
+    "id": "de-dynamic",
+    "language": "de",
+    "voice": "anna",
+    "text": "Franz jagt im komplett verwahrlosten Taxi quer durch Bayern.",
+    "phonemes": "frˈanʦ jˈɑːkt ˈɪm kɔmplˈɛt fɛɾvˈɑːɾlɔstən tˈaksiː kvˈeːɾ dˈʊɐç bˈIɜn."
+  }
+]

validation/package-validation.json

@@ -0,0 +1,78 @@
+{
+  "catalogVersion": "1.0.0",
+  "catalogSchemaValidated": true,
+  "onnxRuntimeVersion": "1.29.0",
+  "artifactsVerified": 9,
+  "cases": [
+    {
+      "id": "pl-mateusz",
+      "phonemeCodePoints": 24,
+      "inputTokens": 26,
+      "outputSamples": 49200,
+      "durationSeconds": 2.05,
+      "peakAbsoluteAmplitude": 0.24918872117996216
+    },
+    {
+      "id": "pl-dynamic",
+      "phonemeCodePoints": 42,
+      "inputTokens": 44,
+      "outputSamples": 60000,
+      "durationSeconds": 2.5,
+      "peakAbsoluteAmplitude": 0.25136497616767883
+    },
+    {
+      "id": "de-anna",
+      "phonemeCodePoints": 74,
+      "inputTokens": 76,
+      "outputSamples": 90000,
+      "durationSeconds": 3.75,
+      "peakAbsoluteAmplitude": 0.35915330052375793
+    },
+    {
+      "id": "de-dynamic",
+      "phonemeCodePoints": 69,
+      "inputTokens": 71,
+      "outputSamples": 84000,
+      "durationSeconds": 3.5,
+      "peakAbsoluteAmplitude": 0.3485631048679352
+    }
+  ],
+  "boundaryCases": [
+    {
+      "language": "pl",
+      "phonemeCodePoints": 1,
+      "inputTokens": 3,
+      "outputSamples": 13800
+    },
+    {
+      "language": "pl",
+      "phonemeCodePoints": 128,
+      "inputTokens": 130,
+      "outputSamples": 177600
+    },
+    {
+      "language": "pl",
+      "phonemeCodePoints": 510,
+      "inputTokens": 512,
+      "outputSamples": 616800
+    },
+    {
+      "language": "de",
+      "phonemeCodePoints": 1,
+      "inputTokens": 3,
+      "outputSamples": 16200
+    },
+    {
+      "language": "de",
+      "phonemeCodePoints": 128,
+      "inputTokens": 130,
+      "outputSamples": 79800
+    },
+    {
+      "language": "de",
+      "phonemeCodePoints": 510,
+      "inputTokens": 512,
+      "outputSamples": 307200
+    }
+  ]
+}

validation/web-de.json

@@ -0,0 +1,19 @@
+{
+  "language": "de",
+  "onnxRuntimeWebVersion": "1.29.0",
+  "executionProvider": "wasm",
+  "cases": [
+    {
+      "id": "de-dynamic",
+      "inputTokens": 71,
+      "outputSamples": 84000,
+      "peakAbsoluteAmplitude": 0.35137680172920227
+    },
+    {
+      "id": "max-phoneme-boundary",
+      "inputTokens": 512,
+      "outputSamples": 307200,
+      "peakAbsoluteAmplitude": 0.32699844241142273
+    }
+  ]
+}

validation/web-pl.json

@@ -0,0 +1,19 @@
+{
+  "language": "pl",
+  "onnxRuntimeWebVersion": "1.29.0",
+  "executionProvider": "wasm",
+  "cases": [
+    {
+      "id": "pl-dynamic",
+      "inputTokens": 44,
+      "outputSamples": 60000,
+      "peakAbsoluteAmplitude": 0.2551245093345642
+    },
+    {
+      "id": "max-phoneme-boundary",
+      "inputTokens": 512,
+      "outputSamples": 616800,
+      "peakAbsoluteAmplitude": 0.16947251558303833
+    }
+  ]
+}

voices/de/df_anna.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d583ccff3cdca2f7fae535cb998ac07e9fcb90f09737b9a41fa2734ec44a8f0b
+size 522240

voices/pl/pm_mateusz.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dc8f2919ede945e6962310b5204b912e64689d5698c571822bf75ad704870cb4
+size 522240