User Feedback: LTX 2.5 Struggles with Lipsync and Talking Photo Quality
No-Vehicle-3508 · reddit · 2026-08-23
A user reports issues when using LTX 2.5 for image + speech audio to video generation (Lipsync/TalkingPhoto). They observed a degradation in the natural motion of the mouth and teeth, and noted that talkingphoto LoRAs from LTX 2.3 do not work well with LTX 2.5. The user seeks insights or solutions from the community regarding whether the model is simply inadequate for this task.
More from Multimodal
- Halcyon: no text pretraining, beats LlamaGen-XL/8 with 18× less compute — JosephJacks_ · 2026-10-16
- Halcyon early-fusion image models match late-fusion with 275x less pretraining compute, open-sourced — davidwromero · 2026-10-16
- Running Wan2.1 14B FP16 on a 12GB laptop GPU: VRAM swap stress test finishes in 4m11s — RYuu-1120 · 2026-10-16
- Striking Midjourney generations: black holes and planetary aesthetics — ciguleva · 2026-10-16
- Fable 5.5 may drop today as Andrew Curran hints at early release — kimmonismus · 2026-10-16
- Kling 4.0 Adds Timeline Editing, Endorsed by 'Seoul Spring' Director Kim Seong-su — xiaohu · 2026-10-16