Sawfwair/LTX-2.5-Full-BF16-MLX

LTX 2.5 Full BF16 for mere.run

This is the self-contained, inference-only Full/Dev LTX 2.5 package used by mere.run on Apple Silicon. It is derived from the immutable Lightricks/LTX-2.5 revision dd53cc2cd45bbeaa3563dfb575cba3f49cf44761.

The dev and distilled video transformers and Gemma 4 text encoder remain BF16. Transformer tensor payload bytes are copied into mere.run's native module-key namespace and physical order. Installation therefore does not retain two official transformers beside generated native optimization copies.

Contents

• native BF16 dev and distilled transformers plus one shared connector;
• official BF16 Gemma 4 text encoder and LTX projection;
• convolutional and diffusion video VAEs plus the audio VAE/vocoder;
• spatial and temporal upsamplers, distilled LoRA, and duration head;
• governing LTX and Gemma license and notice files.

There are no ComfyUI INT8, NVFP4, duplicate source-transformer, IC-LoRA adapter, or training artifacts in this distribution. Optional DFR, HDR, and Dub-It adapters remain separate installs.

Use

mere.run model pull video-ltx25-full-bf16 --accept-model-license
mere.run video generate \
  "an intricate clockwork observatory turns beneath the stars" \
  --model video-ltx25-full-bf16 \
  --output-mode audio-video \
  --output observatory.mp4

No local conversion, re-keying, or mere.run model optimize step is required.

Terms

LTX 2.5 is governed by the LTX-2 Community License Agreement and Acceptable Use Policy included in this repository. The bundled Gemma 4 weights are governed by the included Gemma terms. Review all terms before use or redistribution.