dgrauet/ltx-2.5-mlx-q8

Int8 quantization (group_size 64, transformer block and text-encoder Linear weights) of dgrauet/ltx-2.5-mlx, the MLX conversion of Lightricks/LTX-2.5.

Quantized with mlx-forge (mlx-forge convert ltx-2.5 --quantize --bits 8).

• Transformer variants: dev, distilled

License

These weights are a derivative work and are distributed under the same terms as the model they were converted from. A verbatim copy ships with them as LICENSE, and the upstream text is at https://github.com/Lightricks/LTX-2/blob/main/LICENSE-2_x. Use of these weights is subject to it, including any restrictions it attaches.

Related Projects

• bf16 variant: https://huggingface.co/dgrauet/ltx-2.5-mlx
• q4 variant: https://huggingface.co/dgrauet/ltx-2.5-mlx-q4

LoRAs

• ltx-2.5-22b-distilled-lora-450-bf16.safetensors

Files

• LICENSE (33.63 KB)
• audio_vae.safetensors (101.58 MB)
• chat_template.jinja (17.06 KB)
• connector.safetensors (3.76 GB)
• duration_head.safetensors (3.63 MB)
• embedded_config.json (2.50 KB)
• generation_config.json (255.00 B)
• ltx-2.5-22b-distilled-lora-450-bf16.safetensors (8.29 GB)
• processor_config.json (1.35 KB)
• quantize_config.json (213.00 B)
• spatial_upscaler_x2_v1_0.safetensors (949.62 MB)
• spatial_upscaler_x2_v1_0_config.json (275.00 B)
• split_model.json (964.00 B)
• temporal_upscaler_x2_v1_0.safetensors (249.81 MB)
• temporal_upscaler_x2_v1_0_config.json (273.00 B)
• text_encoder.safetensors (14.91 GB)
• text_encoder_config.json (4.27 KB)
• tokenizer.json (30.68 MB)
• tokenizer_config.json (2.68 KB)
• transformer-dev.safetensors (19.18 GB)
• transformer-distilled.safetensors (19.18 GB)
• vae_decoder_av.safetensors (795.66 MB)
• vae_decoder_conv.safetensors (776.63 MB)
• vae_encoder_av.safetensors (608.34 MB)
• vae_encoder_conv.safetensors (608.34 MB)
• vocoder.safetensors (246.35 MB)