bytedance/seedvr/upscale/video
Seedvr Upscale Video by Bytedance - AI-powered video editing and transformation. Apply style transfer, motion control, lip-sync, and visual effects to existing videos with natural language instructions.
Example output — click Run to generate your own
Send your first request
OpenAI-compatible endpoint with unified authentication and usage tracking.
https://api.sandbase.ai/v1/runbytedance/seedvr/upscale/videoInput Schema
9 parameters · 0 required · 9 optional
| Parameter | Type | Required | Description |
|---|---|---|---|
seed | integer | Optional | The random seed used for the generation process. |
video | string | Optional | The input video to be processed |
noise_scale | number | Optional | The noise scale to use for the generation process. · Min: 0 · Max: 1 · Default: 0.1 |
upscale_mode | string | Optional | The mode to use for the upscale. If 'target', the upscale factor will be calculated based on the target resolution. If 'factor', the upscale factor will be used directly. · Options: target, factor · Default: "factor" targetfactor |
output_format | string | Optional | The format of the output video. · Options: X264 (.mp4), VP9 (.webm), PRORES4444 (.mov), GIF (.gif) · Default: "X264 (.mp4)" X264 (.mp4)VP9 (.webm)PRORES4444 (.mov)GIF (.gif) |
output_quality | string | Optional | The quality of the output video. · Options: low, medium, high, maximum · Default: "high" lowmediumhighmaximum |
upscale_factor | number | Optional | Upscaling factor to be used. Will multiply the dimensions with this factor when `upscale_mode` is `factor`. · Min: 1 · Max: 10 · Default: 2 |
output_write_mode | string | Optional | The write mode of the output video. · Options: fast, balanced, small · Default: "balanced" fastbalancedsmall |
target_resolution | string | Optional | The target resolution to upscale to when `upscale_mode` is `target`. · Options: 720p, 1080p, 1440p, 2160p · Default: "1080p" 720p1080p1440p2160p |
Output Schema
| Field | Type | Description |
|---|---|---|
id | string | Unique identifier for the generation task |
status | string | Task status: pending, running, completed, failed, timeout |
model | string | Model used for the generation |
outputs | array | Array of output items |
outputs[].url | string | URL of the generated artifact |
outputs[].content_type | string | MIME type (e.g. image/png, video/mp4) |
error | object | null | Error details if failed, null on success |
error.type | string | Machine-readable error type code |
error.message | string | Human-readable error description |
Async Workflow
This model uses asynchronous execution. Submit a request and poll for the result.
- Submit — POST to /v1/run, receive an
id - Poll — GET /v1/run/{id} until status is
completed,failed, ortimeout - Retrieve — Read
outputsfrom the completed response
Code Examples
Ready-to-run snippets
# Step 1: Submit
curl -X POST https://api.sandbase.ai/v1/run \
-H "Content-Type: application/json" \
-H "Authorization: Bearer YOUR_API_KEY" \
-d '{
"model": "bytedance/seedvr/upscale/video",
"video": "https://static.sandbase.ai/examples/bytedance/seedvr/upscale/video/input_video_0.mp4",
"noise_scale": 0.1,
"upscale_mode": "factor",
"output_format": "X264 (.mp4)",
"output_quality": "high",
"upscale_factor": 2,
"output_write_mode": "balanced",
"target_resolution": "1080p",
"prompt": "a beautiful sunset over mountains"
}'
# Step 2: Poll result (replace <id>)
curl https://api.sandbase.ai/v1/run/<id> \
-H "Authorization: Bearer YOUR_API_KEY"API README
SeedVR2 Video Upscaler
SeedVR2 Video Upscaler restores and enlarges an existing video while maintaining continuity across its frames. It is designed for footage whose content and timing are already established but whose delivered resolution needs improvement. The output remains a video asset, making the route a post-processing step rather than a new scene generator.
The workflow can target a named resolution or apply a scale factor, then encode the result in one of several delivery formats. Quality and write-mode controls provide a practical balance between visual fidelity, processing profile, and file size. This makes the route adaptable to review copies, web delivery, and higher-resolution masters without changing the underlying sequence.
Highlights
- Temporal consistency during upscaling. The exact video route explicitly emphasizes coherent restoration across frames. This addresses the stability requirement that distinguishes video enhancement from independent still-image processing.
- One-step video restoration. SeedVR2 is described as a one-step large diffusion-transformer model for video restoration. It restores the sequence without relying on an additional diffusion prior.
- Flexible-resolution restoration design. The SeedVR2 project states that the model can restore videos at any resolution. The route turns that capability into target-resolution and scale-factor workflows.
- High-resolution adaptive attention. The published architecture uses adaptive window attention for high-resolution restoration. This supports the model's focus on processing visually demanding video frames within a unified restoration system.
Pricing
| Billing unit | Price |
|---|---|
| Base price per request | $0.001 |
When to Use
| ✅ Good fit | ❌ Consider alternatives |
|---|---|
| Enlarging existing footage while preserving frame continuity | Generating a new video from a written scene |
| Preparing a lower-resolution clip for HD, 2K, or 4K delivery | Processing only one still image |
| Creating a scaled version when a fixed target tier is unnecessary | Requiring motion retiming, new shots, or timeline editing |
| Selecting an output codec for a downstream media pipeline | Needing the original compressed bitstream unchanged |
| Balancing output quality against file-writing priorities | Requiring a custom delivery codec outside the listed formats |
Prompt Guide
This is an input-guided video restoration route and does not use a descriptive text prompt. Supply an accessible video, choose either a target-resolution workflow or a scale-factor workflow, then set the output encoding options.
Video: <publicly accessible video URL>
Upscale mode: <target or factor>
Target resolution or factor: <matching value>
Output format: <X264 mp4, VP9 webm, PRORES4444 mov, or GIF>
Quality/write mode: <desired processing profile>
Example request:
{
"video": "https://example.com/input-video.mp4",
"upscale_mode": "target",
"target_resolution": "1080p",
"output_format": "X264 (.mp4)",
"output_quality": "high",
"output_write_mode": "balanced"
}
Technical Specs
| Spec | Value |
|---|---|
| Model ID | bytedance/seedvr/upscale/video |
| Video input | video URL field |
| Upscale mode | target or factor; default factor |
| Target resolution | 720p, 1080p, 1440p, or 2160p; default 1080p |
| Upscale factor | 1 to 10; default 2 |
| Noise scale | 0 to 1; default 0.1 |
| Output formats | X264 (.mp4), VP9 (.webm), PRORES4444 (.mov), or GIF |
| Output quality | low, medium, high, or maximum |
| Write mode | fast, balanced, or small |
| Output | Video URL and content type |
| Execution | Asynchronous job |
Related
- SeedVR2 Image Upscaler — use the sibling route for a single still image.

