Model card: document PrimeTTS v2 (MB-iSTFT-VITS 34.7M, Xinran, 16 kHz) as flagship; v1 kept as legacy CPU family

README.md

@@ -7,9 +7,12 @@ tags:
   - text-to-speech
   - tts
   - onnx
+  - gguf
   - on-device
   - jetson
   - telephony
+  - vits
+  - mb-istft-vits
   - mandarin
   - taiwanese-mandarin
 base_model: owensong/Inflect-Nano-v1
@@ -18,17 +21,60 @@ library_name: onnxruntime
 pipeline_tag: text-to-speech
 ---
 
-# PrimeTTS — tiny on-device zh-TW + English TTS
+# PrimeTTS — on-device zh-TW + English TTS
 
-A tiny Taiwan-Mandarin + English text-to-speech model that runs **entirely on CPU** (contact-centre, GPS,
-transit). One model, **one young-female voice**: Chinese, English, and code-mix through a single frontend
-(no language routing), built for **entity correctness** — phone numbers, emails, addresses, prices, dates,
-temperatures, %, serials. Two checkpoints share the same acoustic engine:
+Taiwan-Mandarin + English text-to-speech built for on-device use (contact-centre, GPS, transit): one
+voice across Chinese, English, and code-mix through a single frontend (no language routing), with
+**entity correctness** — phone numbers, emails, addresses, prices, dates, temperatures, %, serials.
 
-- **`v1b_16k/` — flagship (~5.0M, 16 kHz)** — the clearest (0–8 kHz band); **default**.
-- **`v1b_8k/` — leanest on-device (4.09M, 8 kHz)** — telephone-band but **RTF 0.35 on a Jetson Nano** (1 thread).
+Two model generations:
 
-> 🔊 **Live demo:** https://huggingface.co/spaces/Luigi/PrimeTTS-vs-Inflect-Nano-v1
+- **`v2_mbistft_16k/` — PrimeTTS v2 (34.7M, 16 kHz) — current flagship.** End-to-end
+  **MB-iSTFT-VITS** targeting the Jetson Nano **GPU** (and any CPU via ONNX). Best quality and
+  intelligibility of the family; female Mandarin voice ("Xinran").
+- **`v1b_16k/` / `v1b_8k/` — PrimeTTS v1 (~5.0M / 4.09M).** FastSpeech + Snake-HiFiGAN, pure-**CPU**,
+  young-female zh-TW voice; `v1b_8k` reaches **RTF 0.35 on a Jetson Nano CPU** (1 thread). Use v1
+  when the deployment budget is CPU-only and tight.
+
+> 🔊 **Live demo (serves v2 + v1):** https://huggingface.co/spaces/Luigi/PrimeTTS-vs-Inflect-Nano-v1
+
+## PrimeTTS v2 (`v2_mbistft_16k/`)
+
+| | PrimeTTS v2 |
+|---|---|
+| **Architecture** | MB-iSTFT-VITS (end-to-end VAE + flow + adversarial; multi-band iSTFT head; conv-only, no LSTM) |
+| **Parameters** | 34.7M (generator) |
+| **Sample rate** | 16 kHz |
+| **Voice** | female Mandarin, "Xinran" |
+| **Eval (36 held-out zh/mix/en sentences)** | X-ASR CER **0.027** overall — zh 0.033 · code-mix 0.039 · en 0.008 (below its 7B teacher's 0.043 on the same eval) |
+| **Runtime** | single ONNX (`primetts_v2_xinran.onnx`, ORT-CPU, RTF ~0.01 on a desktop core) · `primetts_v2_xinran.gguf` for the ggml/CUDA Jetson-Nano runtime (port in progress) |
+
+**Training:** distilled from a **VibeVoice-Large** (MIT) teacher speaking the `zh-Xinran_woman` preset —
+29k utterances over the same entity-rich zh-TW corpus as v1, per-utterance speaker-consistency QC
+(retry-regenerated until >99% of clips match the target voice), trained from scratch at 16 kHz with a
+3-embedding frontend (phone + tone + language, 88 symbols) and deterministic duration predictor.
+
+```python
+# v2 quickstart — one session, one call
+import numpy as np, onnxruntime as ort, soundfile as sf
+import sys; sys.path.insert(0, "PrimeTTS/scripts")
+import frontend_bopomofo as F
+
+sess = ort.InferenceSession("PrimeTTS/v2_mbistft_16k/primetts_v2_xinran.onnx",
+                            providers=["CPUExecutionProvider"])
+o = F.text_to_ids("您好,歡迎使用 PrimeTTS。Thank you for calling.")
+blank = lambda s: np.array([[0] + [v for x in s for v in (x, 0)]], np.int64)  # add_blank=true
+wav = sess.run(None, {"x": blank(o["phone_ids"]), "tone": blank(o["tone_ids"]),
+                      "lang": blank(o["lang_ids"]),
+                      "x_lengths": np.array([2*len(o["phone_ids"])+1], np.int64),
+                      "noise_scale": np.array([0.667], np.float32),
+                      "length_scale": np.array([1.0], np.float32)})[0].reshape(-1)
+sf.write("out.wav", wav, 16000)
+```
+
+---
+
+# PrimeTTS v1 (legacy CPU family)
 
 | | flagship `v1b_16k/` | on-device `v1b_8k/` |
 |---|---|---|
@@ -101,11 +147,13 @@ and a small brand lexicon. Text past `max_frames` is auto-chunked at punctuation
 ## Model files
 
 ```
-v1b_16k/{acoustic_encoder,acoustic_decoder,vocoder}.onnx + meta.json  ← FLAGSHIP (~5.0M, 16 kHz) — the demo serves this
-v1b_8k/ {acoustic_encoder,acoustic_decoder,vocoder}.onnx + meta.json  ← leanest on-device (4.09M, 8 kHz, Jetson Nano)
+v2_mbistft_16k/primetts_v2_xinran.onnx   ← PrimeTTS v2 FLAGSHIP (34.7M, 16 kHz, single ONNX) — the demo serves this
+v2_mbistft_16k/primetts_v2_xinran.gguf   ← same weights for the ggml/CUDA Jetson-Nano runtime
+v1b_16k/{acoustic_encoder,acoustic_decoder,vocoder}.onnx + meta.json  ← v1 16 kHz (~5.0M, CPU)
+v1b_8k/ {acoustic_encoder,acoustic_decoder,vocoder}.onnx + meta.json  ← v1 leanest on-device (4.09M, 8 kHz, Nano CPU)
 {acoustic_encoder,…}.onnx + meta.json   ·   v3_4.6M/                  ← legacy 24 kHz variants (6.85M / 4.63M), for record
 scripts/        frontend, aligner, corpus-gen, train / export, eval
-inflect_nano/   the trainer (acoustic.py + vocoder.py), forked from Inflect-Nano-v1 (LICENSE included)
+inflect_nano/   the v1 trainer (acoustic.py + vocoder.py), forked from Inflect-Nano-v1 (LICENSE included)
 ```
 
 ## Quickstart (CPU)
@@ -172,8 +220,12 @@ own ~10 s reference clip) are in the repo.
 
 ## Credits & licenses
 
-- **Base / trainer:** [`owensong/Inflect-Nano-v1`](https://huggingface.co/owensong/Inflect-Nano-v1) (Apache-2.0)
-- **Teacher:** [`openbmb/VoxCPM2`](https://huggingface.co/openbmb/VoxCPM2) · **Reference voice:**
+- **v2 architecture:** [MB-iSTFT-VITS](https://github.com/MasayaKawamura/MB-iSTFT-VITS) (Kawamura et al., Apache-2.0)
+  · **v2 teacher:** VibeVoice-Large (Microsoft, **MIT**) speaking its `zh-Xinran_woman` preset
+  (via the MIT [community repo](https://github.com/vibevoice-community/VibeVoice)); synthesized speech,
+  AI-generated voice — mark it as such in products
+- **v1 base / trainer:** [`owensong/Inflect-Nano-v1`](https://huggingface.co/owensong/Inflect-Nano-v1) (Apache-2.0)
+- **v1 teacher:** [`openbmb/VoxCPM2`](https://huggingface.co/openbmb/VoxCPM2) · **v1 reference voice:**
   [Mozilla Common Voice zh-TW](https://commonvoice.mozilla.org/datasets) (**CC0 / public domain**)
 - **Gate ASR:** Breeze-ASR-25 (MediaTek Research) · Whisper-medium · **Aligner:**
   `facebook/wav2vec2-lv-60-espeak-cv-ft` + `torchaudio.forced_align` · **Eval:** sherpa-onnx X-ASR