Tasks and inputs
ByteDance-Seed/SeedVR2-3B is listed for Video-to-video, Video Restoration. It accepts video and produces video. Use the publisher’s input format and preprocessing for your chosen deployment.
SyncDev model directory · Profile updated Oct 7, 2026
SeedVR2-3B is a ByteDance Seed video-restoration checkpoint. Compare temporal fidelity, runtime memory and prototype limitations before deployment.
The specification score summarizes recorded features and limits. It is not a benchmark result.
Recorded modality support shows what this model can accept and produce. An undocumented modality is shown as unknown rather than assumed to be unsupported.
Accepts
Produces
SeedVR2-3B is ByteDance Seed's open-weight checkpoint for restoring existing video with a one-step diffusion approach. It is a restoration model, rather than a general text-to-video generator.
Use footage that resembles your actual inputs: moving subjects, faces, text, compressed backgrounds and changes in lighting. Inspect temporal stability and unwanted invented detail, not only the sharpness of one frame.
The publisher describes these releases as prototypes. Heavy degradation and large motion can produce unreliable restoration. Lightly degraded material can become oversharpened, so an enhancement is not automatically a more faithful reconstruction.
Compare this release with SeedVR2-7B on identical clips and output settings. Record GPU memory use, elapsed time, precision and processing options from the runtime you actually use. A parameter count alone cannot supply a reliable VRAM minimum or speed estimate.
Weights are under Apache 2.0. Use the project's supported inference workflow and the correct checkpoint. A Hugging Face download is not a managed API; this profile does not infer a token price or hosting offer. Browse video restoration models for related checkpoints.
Compare the model on the work you need to complete.
ByteDance-Seed/SeedVR2-3B is listed for Video-to-video, Video Restoration. It accepts video and produces video. Use the publisher’s input format and preprocessing for your chosen deployment.
Choose duration, resolution and frame rate together. Compare the cost of an accepted clip, including repeat generations.
The weights are downloadable under Apache-2.0. Check runtime support and size the complete model and working context for your expected concurrency.
These records share a directory category. Check the source card, output type and deployment requirements before treating a model as a replacement.
ByteDance-Seed
SeedVR2-7B is a ByteDance Seed video-restoration checkpoint. Compare temporal fidelity, runtime memory and prototype limitations before deployment.