Sawfwair/LTX-2.5-Distilled-BF16-MLX-Q4-Text

LTX 2.5 Distilled BF16 for mere.run

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

The video transformer remains BF16. Its tensor payload bytes are copied into mere.run's native module-key namespace and physical order so installation does not retain an upstream checkpoint beside a generated optimization cache. The Gemma 4 language tower uses MLX affine Q4/group-64 for eligible language weights; the LTX projection and tokenizer assets remain BF16/raw.

Contents

• native BF16 distilled transformer and shared connector;
• MLX Q4 Gemma 4 text tower with BF16 LTX projection;
• convolutional video VAE, audio VAE/vocoder, spatial upsampler, and duration head;
• governing LTX and Gemma license and notice files.

There are no ComfyUI INT8, NVFP4, dev-transformer, DiffVAE, temporal-upscaler, or training artifacts in this distribution.

Use

mere.run model pull video-ltx25-distilled-bf16 --accept-model-license
mere.run video generate \
  "a clockwork observatory turns beneath a starry sky" \
  --model video-ltx25-distilled-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.