Help Needed: Wan 2.2 Image-to-Video Upscaling Workflow
Far-Solid3188 · reddit · 2026-07-06
While using Wan 2.2's 14B fp8 i2v model for image-to-video generation on an RTX 5090, a creator encountered issues with over-smoothed visuals and loss of detail. They tried using Svr2 with latent noise adjustments to restore details, but without success. They are currently waiting to download a community FP32 version to try out, and are asking for the latest best practices for Wan 2.2 upscaling workflows, noting that most available tutorials are outdated from last year.
More from Multimodal
- Dense instance segmentation via Muse Wonder + SAM 3.1 on Meta's Model API — nikhilaravi · 2026-10-26
- LichtFeld Studio v0.5.4.3 ships community-built SteadySplats, POPSpa and Voxelization — janusch_patas · 2026-10-26
- One prompt to a 2-3 min narrated short: hands-on test of Antigravity video generation — AI_Andrew · 2026-10-26
- Creator makes 10,564-frame animated short entirely with Gemini toolchain — AI_Andrew · 2026-10-26
- FLUX.3 Klein step-distilled to 4-8 steps with CFG baked in — Fit-Structure-6489 · 2026-10-26
- ostris pre-training "Primal VAE": 1-channel latents match modern 8x VAEs via patching — ostrisai · 2026-10-26