Tools / Video Upscale
Video Upscale icon

Video Upscale

AI video upscaling up to 10x

Upscale video resolution to 1080p, 2K, or 4K with multiple AI models. Choose general-purpose, premium face/texture recovery, professional noise reduction, or budget upscaling.

2 skillsv0.01
Upscale Video

Upscale a video up to 10x resolution or to a target resolution (1080p, 2K, 4K). Choose a model optimized for the content type — general footage, faces, AI-generated, or old film.

Returns: Upscaled video URL, downloadable asset, model used, and cost
Parameters
video_url *stringURL of the video to upscale. The tool server CAN fetch any public URL — always pass it.
modelstringUpscale model to use. Default: bytedance.
upscale_factornumberScale factor (1-10 for SeedVR2/Bytedance, 1-4 for Topaz/FlashVSR, 1-8 for RealESRGAN). Default 2x. Ignored if target_resolution is set.
target_resolutionstringTarget output resolution. Not all models support all resolutions.
enhancement_presetstringContent-aware enhancement preset. Bytedance only. Default: general.
enhancement_tierstringEnhancement quality tier. Bytedance only. Pro is 10x cost. Default: standard.
fidelitystringFidelity level. Bytedance only. Default: high.
noise_scalenumberNoise control (0-1). SeedVR2 only.
output_formatstringOutput video format. SeedVR2 and FlashVSR only. Default: X264 (.mp4).
output_qualitystringOutput encoding quality. SeedVR2 and FlashVSR only. Default: high.
accelerationstringProcessing speed vs quality tradeoff. FlashVSR only. Default: regular.
qualitynumberTile blending quality (0-100). FlashVSR only. Default: 70.
color_fixbooleanEnable color correction. FlashVSR only. Default: true.
preserve_audiobooleanPreserve original audio track. FlashVSR only. Default: false.
topaz_variantstringTopaz AI model variant. Topaz only. Default: Proteus.
target_fpsnumberTarget frame rate (16-60). Enables frame interpolation. Bytedance and Topaz only.
compressionnumberCompression artifact removal (0-1). Topaz only.
noisenumberNoise reduction (0-1). Topaz only.
halonumberHalo reduction (0-1). Topaz only.
grainnumberFilm grain amount (0-1). Topaz only.
recover_detailnumberDetail recovery (0-1). Topaz only.
h264_outputbooleanOutput H264 codec instead of H265. Topaz only. Default: false.
seednumberRandom seed for reproducible results. SeedVR2 and FlashVSR only.
Example
Upscale a video to 1080p
curl -H "Authorization: Bearer $TOOLROUTER_API_KEY" \
  -d '{
  "tool": "video-upscale",
  "skill": "upscale_video",
  "input": {
    "video_url": "https://example.com/video.mp4"
  }
}' \
  https://api.toolrouter.com/v1/tools/call
List Models

List all available video upscale models with key, pricing, max upscale factor, description, and variants.

Returns: All available video upscale models with key, name, pricing, max factor, description, and variants
Example
List all video upscale models
curl -H "Authorization: Bearer $TOOLROUTER_API_KEY" \
  -d '{
  "tool": "video-upscale",
  "skill": "list_models",
  "input": {}
}' \
  https://api.toolrouter.com/v1/tools/call
Loading reviews...
Loading activity...
v0.012026-03-23
  • Initial release: 5 upscale models (15+ variants), target resolution and factor modes

Quick Start

MCP (Claude Code)
claude mcp add --transport stdio \
  --env TOOLROUTER_API_KEY=YOUR_API_KEY \
  toolrouter -- npx -y toolrouter-mcp
REST API
curl -H "Authorization: Bearer $TOOLROUTER_API_KEY" \
  -d '{"tool":"video-upscale","skill":"upscale_video","input":{}}' \
  https://api.toolrouter.com/v1/tools/call

Frequently Asked Questions

Can it upscale a clip to 1080p, 2K, or 4K?

Yes. `upscale_video` can target common HD and 4K outputs or scale up by a factor when you need a bigger master.

How do I choose the right model?

`list_models` shows the available video upscalers so you can pick one for general footage, faces, AI-generated clips, or old film.

What kind of footage works best?

The tool is tuned for regular footage, face-heavy shots, AI-generated video, and older material that needs noise reduction and texture recovery.

What comes back after the job finishes?

You get a downloadable upscaled video plus the model used, so it is easy to compare results or rerun the job with a different setting.