Create video and synchronized audio from text or a starting frame. Use LTX 2.3 with motion-tracking and director graphs, or opt into LTX 2.5's updated prompt understanding and distilled model.
ComfyUI and JupyterLab with LTX 2.3 enabled by default
Text/image-to-video, motion tracking, Face-ID and Director graphs
Optional LTX 2.5 text-to-video and image-to-video graphs
One combined text/image-to-video graph, motion tracking, Face-ID and Director.
LTX 2.5 2 workflows · Optional
Separate text-to-video and image-to-video graphs. Added only after the access check passes.
Before you deploy
LTX 2.3 is already enabled; leave it unchanged to use that model family. In Edit Template, open Environment Variables. For LTX 2.5, change download_ltx25 from false to true and add the model access token below before deploying.
Model selection
download_ltx23Default: true
LTX 2.3 is already enabled. Leave unchanged, or set false if you do not want the 2.3 set.
download_ltx25Default: false
Change false to true for LTX 2.5. Accept its model licence and add HF_TOKEN before deploying; failed access checks skip its models and workflows.
disable_ic_lorasDefault: false
Change to true to skip the IC LoRA collection. This switch is independent of the model-family flags; gated files also need an authorised HF_TOKEN.
Model access
HF_TOKENDefault: empty
Hugging Face read token from an account that has accepted the licences for LTX 2.5 and any gated IC LoRAs you want.
CivitAI downloads
civitai_tokenDefault: empty
Your CivitAI API token. Add it if you want to download your own LoRAs or checkpoints.
CIVITAI_LORASDefault: empty
Comma-separated LoRA model-version IDs to download into models/loras. Replace any replace_with_ids placeholder with your IDs.
CIVITAI_CHECKPOINTSDefault: empty
Comma-separated checkpoint model-version IDs to download into models/checkpoints. Leave empty if you do not need any.
