A Reddit user builds a Wan 2.2 continuation workflow with LoRAs and ComfyUI
SnooMacaroons1365 · reddit · 2026-07-28
A Reddit user shows their first longer Wan 2.2 continuation-generation clip, built from a chain of short I2V shots and joined in DaVinci Resolve. They say the workflow uses Wan 2.2 I2V Hi/Lo Noise, lightx2v4step LoRA, a trained character LoRA, and Wan Continuation Conditioning in ComfyUI.
Key details:
- The creator spent a month learning prompt control and model workflows, moving from AUTOMATIC1111 to Flux and then Wan.
- They trained a character LoRA from a Flux-generated dataset and are still polishing realism details such as skin texture.
- On a 16 GB VRAM 5070 Ti, each 2-second shot at 30 fps took 3–5 minutes to generate; full 14B without the 4-step setup took 10–15 minutes per 81 frames at 16 fps.
- The user notes some stutters and patchy head movement, and asks for feedback.
The attached image shows one of the generated character frames, reinforcing that this is a practical image/video generation workflow post rather than a pure announcement.
More from Multimodal
- Spanish-Language AI Media Generation Video Tutorial Series Launched, from Basics to Advanced — Botoni · 2026-07-28
- Leaked Imagine Omni upgrade would bring images, video and voice into Grok Imagine — XFreeze · 2026-07-28
- ACE-Step UI claims to generate 4-minute vocal songs locally, free and open source — anselm · 2026-07-28
- User shares a new Midjourney style with exact prompt settings — azed_ai · 2026-07-28
- A 5 MB McBess-style LoRA for Krea2 trained on 120 captioned images — Winter_unmuted · 2026-07-28
- Topview launches Film Studio with 3D blocking and micro-expression controls — XFreeze · 2026-07-28