t2v models are not visible under video in the gallery. Is this expected?

[gadeynebram]

t2v models are not visible under video in the gallert. Is this expected?

I'm currently downloading ltx-2.3-22b-distilled-ggml but it is listed under image? Another example is wan-2.1-t2v-1.3b-ggml

I'm on the latest release 4.3.3

[cananoo]

This was a gallery metadata issue in the 4.3.3 timeframe, rather than a limitation of the models.

The Wan 2.1 T2V entry uses the gallery template gallery/wan-ggml.yaml. That template was missing the known_usecases: video value, so the UI could classify the model in the wrong section even though its tags and backend were for video generation. The metadata was fixed in the merged PR #11214 by adding the video use case.

The LTX-2.3 entries use gallery/ltx-ggml.yaml, which declares known_usecases: video; the current gallery index also tags them as text-to-video, image-to-video, and video-generation.

The fix is included in LocalAI v4.8.0 and later, including v4.9.0. If you are still running v4.3.3, upgrade to a newer release and refresh the gallery data. If a current release still shows the model under Image, check that the instance is using the current gallery index rather than a cached or custom gallery URL.

References:

• Fix use case for video model #11214
• https://github.com/mudler/LocalAI/releases/tag/v4.9.0
• https://localai.io/docs/features/model-gallery/