Help: Wan 2.2 Local Settings for Realistic Human Video Generation
Kind-Illustrator6341 · reddit · 2026-07-05
A user shared their experience generating realistic human face/body videos locally using Wan 2.2 (ComfyUI + RunPod), noting Wan 2.2 outperforms LTX on local tasks. Despite inputting resolutions of 2048x2048 or 1792x2400, actual generation resolutions drop to 1024x2024 or 832x1216, and they are unsure how to balance quality.
Using an INT8 model on an RTX 4090 takes 20 minutes per generation, which feels too slow. They are seeking advice on: (1) the optimal settings to maintain image quality, and (2) which INT8-compatible RunPod GPU to choose to achieve generation in under 5 minutes.
More from Multimodal
- TischEins, a large Midjourney SREF library, relaunches with new design after security outage — tisch_eins · 2026-10-26
- MOSS-Image-SFT 4-step distillation tops 16.5K downloads on Hugging Face — _akhaliq · 2026-10-26
- Claude Code can now edit videos: a 10-prompt workflow from raw footage to final cut — TawohAwa · 2026-10-26
- One-sentence 3D worlds turn cinematic with CineMap and TapNow — gizakdag · 2026-10-26
- No Camera Needed: Creator Builds an Entire World in CineMap and Directs It Into Cinema — gizakdag · 2026-10-26
- Kling vs. Seedance: the Chinese AI video model rivalry — DavidmComfort · 2026-10-26