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://huggingface.co/Lightricks/LTX-2.3/blob/main/LICENSE. Use of these weights is subject to it, including any restrictions it attaches.

Usage

These weights can be used with ltx-2-mlx, a native MLX inference pipeline for LTX-2.3 on Apple Silicon:

bash

git clone https://github.com/dgrauet/ltx-2-mlx.git && cd ltx-2-mlx
uv sync --all-extras

ltx-2-mlx generate -p "A sunset over the ocean" --distilled \
    --model dgrauet/ltx-2.3-mlx-q4 -o sunset.mp4

Keep quantize_config.json next to the weights (the loader also infers bits/group_size from the weight shapes if it is missing).

Related Projects

Code: https://github.com/dgrauet/ltx-2-mlx
ComfyUI compatible custom nodes: https://github.com/dgrauet/ComfyUI-LTXVideo-mlx
bf16 variant: https://huggingface.co/dgrauet/ltx-2.3-mlx
q8 variant: https://huggingface.co/dgrauet/ltx-2.3-mlx-q8

LoRAs

ltx-2.3-22b-distilled-lora-384-1.1.safetensors
ltx-2.3-22b-distilled-lora-384.safetensors

Files

LICENSE (20.90 KB)
audio_vae.safetensors (101.57 MB)
config.json (951.00 B)
connector.safetensors (5.91 GB)
embedded_config.json (7.06 KB)
ltx-2.3-22b-distilled-lora-384-1.1.safetensors (7.08 GB)
ltx-2.3-22b-distilled-lora-384.safetensors (7.08 GB)
quantize_config.json (100.00 B)
spatial_upscaler_x1_5_v1_0.safetensors (1.02 GB)
spatial_upscaler_x1_5_v1_0_config.json (274.00 B)
spatial_upscaler_x2_v1_1.safetensors (949.62 MB)
spatial_upscaler_x2_v1_1_config.json (275.00 B)
split_model.json (1.56 KB)
temporal_upscaler_x2_v1_0.safetensors (249.81 MB)
temporal_upscaler_x2_v1_0_config.json (273.00 B)
transformer-dev.safetensors (10.54 GB)
transformer-distilled-1.1.safetensors (10.54 GB)
transformer-distilled.safetensors (10.54 GB)
vae_decoder.safetensors (776.62 MB)
vae_encoder.safetensors (608.33 MB)
vocoder.safetensors (246.35 MB)