Models listed
2
Published in this category
Worth weighing up
What goes in and out, licence terms, and whether you can host it yourself
How we treat data
Published specs only — we don't estimate numbers a vendor hasn't stated
Video restoration models improve existing footage rather than generate a new clip solely from a text prompt. They are relevant to denoising, restoring degraded frames and related enhancement workflows, subject to each checkpoint's documented scope.
Inspect motion, flicker, faces, text and consistency across the whole clip. Restoration can invent detail, so compare outputs with the source instead of assuming that a sharper result is more faithful. Record resolution, clip length, precision and runtime alongside processing time.
Compare SeedVR2-3B and SeedVR2-7B on the same footage before choosing a checkpoint.
Scan the table below, then open any model for pricing, limits and the full write-up.
| ByteDance-Seed/SeedVR2-3B ByteDance-Seed | 10.0 | — | — | — | — |
| ByteDance-Seed/SeedVR2-7B ByteDance-Seed | 10.0 | — | — | — | — |
1–2 of 2 models · select a column to sort
About this order: Recommended shows our editorial shortlist first, followed by recent releases. The spec score is a 0–100 figure SyncDev calculates from each vendor's published specifications — context window, reasoning support, input modalities, tool calling, maximum output and how recently the model shipped. It measures breadth of capability, not benchmark performance, so a higher-scoring model is not automatically the better choice for your task.