What does SeedVR2 document for video restoration (one-step diffusion, GPU needs, scripts) and where does it say it fails or over-sharpens?
Problem. Restoring or upscaling AI-generated or degraded video without temporal artefacts. Conditions. SeedVR / SeedVR2 (ByteDance Seed; ICLR 2026), Apache 2.0; 3B/7B models. Method. SeedVR2 is a one-step diffusion video restoration model with adaptive window attention that adjusts window size to the output resolution. Inference: torchrun projects/inference_seedvr2_3b.py with --video_path, --output_dir, --seed, --res_h, --res_w and --sp_size (sequence parallel GPUs). Stated capacity: one H100-80G handles a 100-frame 720x1280 video; four H100-80G support 1080p and 2K with sp_size=4; tiling tricks were not yet supported. A community ComfyUI node exists (linked in README). Failure modes (README Limitations). Prototype models; not robust to heavy degradation or very large motions; may fail to remove degradation or generate unpleasing detail; on lightly degraded inputs such as 720p AIGC video it may over-generate detail and oversharpen, especially at small resolutions like 480p. Method (...
Structured fields
{
"citations": [
{
"url": "https://arxiv.org/abs/2506.05301",
"note": "Abstract; Inference; GPU Requirement; GPU Requirement; Limitations; Limitations; License; finishing order"
},
{
"url": "https://raw.githubusercontent.com/ByteDance-Seed/SeedVR/main/readme.md",
"note": "Abstract; Inference; GPU Requirement; GPU Requirement; Limitations; Limitations; License; finishing order"
}
]
}Evidence claims
- [sourced_fact] Problem. Restoring or upscaling AI-generated or degraded video without temporal artefacts. Conditions. SeedVR / SeedVR2 (ByteDance Seed; ICLR 2026), Apache 2.0; 3B/7B models. Method. SeedVR2 is a one-step diffusion video restoration model with adaptive window attention that adjusts window size to the output resolution. Inference: torchrun projects/inference_seedvr2_3b.py with --video_path, --output_dir, --seed, --res_h, --res_w and --sp_size (sequence parallel GPUs). Stated capacity: one H100-80G handles a 100-frame 720x1280 video; four H100-80G support 1080p and 2K with sp_size=4; tiling tricks were not yet supported. A community ComfyUI node exists (linked in README). Failure modes (README Limitations). Prototype models; not robust to heavy degradation or very large motions; may fail to remove degradation or generate unpleasing detail; on lightly degraded inputs such as 720p AIGC video it may over-generate detail and oversharpen, especially at small resolutions like 480p. Method (inference). Do not run it on clean generated footage that is already sharp; reserve it for genuinely degraded sources; compare against the original at 100% zoom for hallucinated texture on faces, text and fine patterns; run it on the final cut only (after edits, lip sync and colour). Limitations. README claims only; no independent evaluation read. Version/date. README (2026), arXiv 2506.05301; fetched 2026-10-23. · locator: Abstract; Inference; GPU Requirement; GPU Requirement; Limitations; Limitations; License; finishing order source
Rights
| component | license | hosting | redistribution | modification | training | Evidence and notes |
|---|---|---|---|---|---|---|
| documentation | unknown | permitted | permitted | unknown | unknown | Applies to this original summary only, not to the cited sources. |