# ComfyUI character LoRAs into an edited Seedance 2.5 fight — retained record

> Retained reddit record from reddit:comfyui.

Canonical: https://highsignal.app/data/records/b7b26bfbf763bf34

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:comfyui)  published 2026-11-14 18:17 ingested 2026-11-15 02:36

[original source ↗](https://www.reddit.com/r/comfyui/comments/1xw99cs/comfyui_character_loras_into_an_edited_seedance/)

## retained text



```
I created Sunny and Sora in ComfyUI with my custom Z-Image LoRAs and character sheets. That was the character/reference stage; the fight footage was generated with Seedance 2.5 through Higgsfield.



I wanted Sunny to look capable of winning early, before Sora gains the upper hand. Seven generated takes were inspected. The strongest take became the backbone, with selected material from the original 15-second take, Close Combat and comparisons A and B. Four impact clips I cut from upscaled footage were all incorporated.



The first assembly was in Blender, then I refined the edit in DaVinci Resolve with ChatGPT/Codex helping on prompting, review and fixes. I cut an awkward reset, added a short bridge toward the doorway, and worked around a garbled transition. Continuity mattered more than including every attractive shot.



Finished in local Topaz Video AI at 2560×1440, retaining the approved 60 fps treatment. The original instrumental, Corridor Pressure, was made in Suno V6 and fitted to the picture while keeping the fight effects present.



This is fictional AI-generated action.
```



document fetched 2026-11-15 00:36 · canonical https://reddit.com/r/comfyui/comments/1xw99cs/comfyui_character_loras_into_an_edited_seedance
