ComfyUI Newbie Bug: Swapping LoRAs Yields the Exact Same Video Output
Ndsis2 · reddit · 2026-09-19
A newcomer moving from A1111 to ComfyUI Desktop reports a puzzling issue:
- The standard T2V template generates video normally, but adding or swapping a LoRA and re-running with the same prompt in the same tab produces a completely identical video—no change at all.
- Restarting ComfyUI, or sometimes opening a new tab, is the only way to get a different result with the new LoRA.
- In other words, only changing the LoRA (or even re-running unchanged) yields byte-identical output; the workflow seems not to re-execute properly. The poster is asking what they're doing wrong.
For ComfyUI video users, it's a reminder to watch caching/seed behavior after changing LoRAs and re-run in a fresh state when needed.
More from Multimodal
- Scoop: Runway is working on a new AI model with early access open — tlakomy · 2026-09-19
- Midjourney v8.2 Prompt Recipe for Kodak Film-Style Stereoscopic Portraits — michaelrabone · 2026-09-19
- Forcing Wan to Generate a 60-Second Single Shot on a 16GB GPU: It Finished, Barely — Wonderful_Sample6291 · 2026-09-19
- Reddit asks: models for voice reconstruction and remaster of low-quality audio? — Ant_6431 · 2026-09-19
- Aethr Create launches: one studio aggregating 50+ AI models with free daily generations for artists — koltregaskes · 2026-09-19
- Better video generation: use crafted image references, says Umesh — umesh_ai · 2026-09-19