ReCaVSR ported to ComfyUI: one-step diffusion video upscaling hits 4K in 36s on RTX 4090
LatentSpacer · reddit · 2026-11-11
A developer ported ReCaVSR — a one-step streaming video super-resolution model from USTC built on Wan2.2 TI2V-5B — to native ComfyUI nodes, open-sourced on GitHub.
Key specs
- 4× or 2× upscaling with one diffusion step per 8-frame block; FlashDecoder streams decoding so longer clips cost time, not VRAM
- 149 frames 960×480 → 3840×1920 in 36s on RTX 4090, 67s on 3090
- int8 ConvRot transformer: 1.8× faster, 4 GB less VRAM, near-bf16 quality
- 1440×1440 output runs on an 8 GB card via weight streaming
- Update: arbitrary-length videos, up to 5.5K output on 24 GB cards, auto chunking with KV-cache offload and NVENC encoding
Note it's a diffusion upscaler — fine detail is generated, not recovered. Installable via ComfyUI-Manager; int8/bf16 weights on Hugging Face.
Related event: ReCaVSR Ported to ComfyUI for One-Step 4K Video Super-Resolution(3 posts)→
More from Multimodal
- AI video model nails multi-shot continuity as users stitch random ideas into clips — HeyZoyaKhan · 2026-11-11
- Amazon Prime 360 uses AI to rebuild NBA plays in 3D for live, any-angle replays — import_jmr · 2026-11-11
- Adobe MAX demo: designer builds a custom Firefly video transition live on stage — umesh_ai · 2026-11-11
- Sophon-Realtime-v1, a full-duplex speech-to-speech model, trends on Hugging Face — adventists-ai · 2026-11-11
- Asked Fable 5.2 to 'go wild' making a game, got a Pong clone — and loved it — ariG23498 · 2026-11-11
- Western town built with Fable 5.2: ray-traced weathering ages every board — ravisparikh · 2026-11-11