Update LTX model library snippets for LTX-2.5 and DFR - #2404
Conversation
![cursor[bot]](https://avatars.githubusercontent.com/in/1210556?s=60&v=4){kind=small}
There was a problem hiding this comment.
Cursor Bugbot has reviewed your changes using default effort and found 1 potential issue.
❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.
Want higher recall? High effort reviews run extra passes and find more bugs. A team admin can switch effort levels in the Cursor dashboard.
Reviewed by Cursor Bugbot for commit b9cdff5. Configure here.
| ..._ltx25SplitArgs(dfr), | ||
| `--detailing-lora ${detailingDir}/${_LTX_DETAILING_LORA_FILE}`, | ||
| "--spatial-upscalings 1", | ||
| "--temporal-upscalings 1", |
There was a problem hiding this comment.
Wrong DFR CLI upscaling flags
High Severity
The LTX-2.5 DFR Hub snippet emits --spatial-upscalings and --temporal-upscalings, but the LTX-2 CLI documents --temporal-upsample-rounds for temporal control and has no matching --spatial-upscalings flag. Copied snippets will fail at argparse with unrecognized arguments on model pages.
Additional Locations (1)
Reviewed by Cursor Bugbot for commit b9cdff5. Configure here.
Wauplin
left a comment
Wauplin
left a comment
There was a problem hiding this comment.
Thanks for the PR!
Made a quick review pass but I'll defer to @apolinario and @linoytsaban to check actual changes.
There was a problem hiding this comment.
I'll let @apolinario @linoytsaban review that part
There was a problem hiding this comment.
Thanks @Wauplin @alexanderar, took a look now, LGTM pending the dfr pipeline merge to ltx main
Keep LTX-2.3 monolith + Gemma 3 paths, and add LTX-2.5 split-pack snippets with distilled and DFR pipelines.
b9cdff5 to
b4ec92e
Compare
|
@Wauplin |
Wauplin
left a comment
Wauplin
left a comment
There was a problem hiding this comment.
Great! Thank you for the iteration! 🤗
Should be live on the Hub within a few days
3 participants
Summary
LTX.iotoLTX--transformer-path, text encoder, VAEs), with distilled + DFR pipelines--gemma-root, distilled + HQ)base_model; cover LoRA and IC-LoRA variants for both generationsTest plan
pnpm --filter tasks lint:checkpnpm --filter tasks checkpnpm --filter tasks format:checkpnpm --filter tasks testLightricks/LTX-2.5andLightricks/LTX-2.3(and a 2.5/2.3 LoRA)Note
Low Risk
Changes only affect displayed install/run snippets and library labeling on model pages, with no runtime or API impact; wrong 2.5 vs 2.3 detection could show misleading commands to users.
Overview
Updates the LTX model library UI label from
LTX.ioto LTX-2 and rewrites theltxHub code snippets so they match both LTX-2.5 and legacy LTX-2.3 layouts.LTX-2.5 models are detected from
model.id, tags, orbase_model(e.g.ltx-2.5). Snippets use split checkpoint paths (--transformer-path, VAEs, upsamplers, bundled Gemma 4 text encoder),uv sync --extra natten, multi-filehf download(plus a separate detailing IC-LoRA repo for DFR), and examples for distilled, DFR, LoRA, and IC-LoRA pipelines. LTX-2.3 keeps monolith checkpoints andgoogle/gemma-3-12b--gemma-rootwith the same tag-based LoRA / IC-LoRA / base flows, now built through shared_ltxRunhelpers and optional image-to-video hints.Reviewed by Cursor Bugbot for commit b4ec92e. Bugbot is set up for automated code reviews on this repo. Configure here.