Video Upscaling API for Apps, Pipelines, and Agents

Send a clip, get it back sharper and smoother. Five super resolution models, up to 4K, frame interpolation to 60 fps in the same pass, and a price per second of output.

Output
Up to 2160p
Frame rate
Up to 60 fps
Clip length
Up to 5 min
curl -X POST https://upsampler.com/api/v2/video/upscale/rtx-vsr \
  -H "Authorization: Bearer $UPSAMPLER_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: $(uuidgen)" \
  -d '{
    "input": {
      "videoUrl": "https://example.com/clip.mp4",
      "upscaleFactor": 3,
      "durationSeconds": 8,
      "width": 1280,
      "height": 720,
      "fps": 60
    }
  }'
200 OK
{
  "id": "cmh2x9q1v0007",
  "status": "IN_QUEUE",
  "creditCost": 7
}

Pick a model

rtx-vsr and gan are fast and the two that reach 4K. ltx-2-5, seed-vr-2, and relay-vsr rebuild more detail, up to about 4 megapixels per frame.

Video upscaling

  • POSTLTX 2.5/api/v2/video/upscale/ltx-2-5

    Rebuilds the most detail; a creativity setting controls how much.

    2 credits per MP·sUp to about 4 MP
  • POSTSeedVR 2/api/v2/video/upscale/seed-vr-2

    Diffusion upscaler, strong on AI clips and heavy compression.

    2.5 credits per MP·sUp to about 4 MP
  • POSTRelayVSR/api/v2/video/upscale/relay-vsr

    Faithful detail with steady motion, nothing invented.

    0.5 credits per MP·sUp to about 4 MP
  • POSTRTX VSR/api/v2/video/upscale/rtx-vsr

    Fast, clean sharpening all the way to 4K.

    0.1 credits per MP·sUp to 4K (2160p)
  • POSTGAN/api/v2/video/upscale/gan

    Fast GAN upscaling to 4K.

    0.2 credits per MP·sUp to 4K (2160p)
  • POSTRecommend/api/v2/video/upscale/recommend

    Measures the clip and returns which model to use, with the exact input and price.

    Free
  • POSTAuto/api/v2/video/upscale/auto

    Runs the recommended model in one call.

    The model's price

How it works

Every job is asynchronous: submit the clip, and the result comes back on its own.

  1. 1

    Send the clip

    POST to /video/upscale/{model} with a public videoUrl, or upload through POST /uploads and submit the returned file URL. Set fps to 30, 48, 50, or 60 to interpolate in the same pass, for 0.01 credits per second for each fps above 24.

  2. 2

    We process

    The job runs on our GPUs. Your app doesn't wait or hold a connection, and a failed job is refunded automatically.

  3. 3

    Get the result

    Pass a webhook and get one signed POST when it finishes, or poll /status/{id}. Result URLs expire after 3 hours; the status endpoint issues fresh ones on demand.

Clips through the same models

Runs from the same models the API serves.

Drag the split to compare. Zoom in to check the detail.

Built for agents

The full reference is published as plain markdown and OpenAPI, generated from the same schemas the API validates with. Start at the API overview.

Docs agents can read
The whole reference as plain markdown, generated from the schemas the API validates with, so it never drifts from what the endpoints accept.
/api/v2/llms.txt
OpenAPI 3.1
Point a code generator or an agent's tool loader at the spec and every endpoint becomes a typed call.
/api/v2/openapi.json
Safe retries
Send an Idempotency-Key with each submission. A retried request returns the original job instead of charging twice.
Idempotency-Key: <uuid>
Results that come back on their own
Pass a webhook and get one signed POST when the job finishes, or poll the status endpoint. Failed jobs are refunded automatically.
"webhook": "https://..."

Prefer a browser? The video upscaler in the app runs the same models, and the image upscaling API does the same for stills.

Pricing

Pay as you go, or subscribe

One credit balance across every tool.

Pay as you go

$0/month
4¢per credit
  • No monthly commitment. Buy credits whenever you need them.

  • Extra credits never expire

  • 2 jobs at a time

  • 25 jobs in queue

Pro

$15/month
1.5¢per credit
  • 1,000 monthly credits, renewed on your billing date

  • Extra credits never expire

  • 4 jobs at a time

  • 50 jobs in queue

Premium

$30/month
1.3¢per credit
  • 2,300 monthly credits, renewed on your billing date

  • Extra credits never expire

  • 6 jobs at a time

  • 75 jobs in queue

Business

$95/month
1.2¢per credit
  • 8,000 monthly credits, renewed on your billing date

  • Extra credits never expire

  • 15 jobs at a time

  • 100 jobs in queue

  • Contact us for customization of concurrent jobs and queue size

Every plan includes full API access and a commercial use license. Monthly credits renew each billing cycle and do not carry over. Extra credits never expire.

How credits work

Credits depend on the tool and the model, charged per image, per megapixel of output, or per second of video.

Frequently Asked Questions

Yes, with the rtx-vsr or gan model, which go up to 2160p. ltx-2-5, seed-vr-2, and relay-vsr add more detail and stop at about 4 megapixels.

Every submission returns a job id. Poll the status endpoint until it reports SUCCESS, or pass a webhook URL and receive one signed POST when the job finishes.

It is refunded automatically, and the error carries a stable code you can branch on.

You pay per second of output video in credits, priced by output size, with no subscription. A target frame rate adds 0.01 credits per second for each fps above 24, so 60 fps adds 0.36 credits per second. Each response includes the job's credit cost.

Up to 5 minutes per job. Longer footage can be split and sent as several jobs.

Ready to build?

Get an API key and send your first clip. Credits never expire, and failed jobs are refunded.

Get an API key