AI video upscaler API fields: factor 1.1 to 4 and fast, standard, pro

The Sume Video Upscale request has one required field, video_url, and four optional ones that matter: upscale_factor from 1.1 to 4, scale_ratio with the same range, duration_seconds from 1 to 30, and enhancement_tier set to fast, standard or pro. The public rate is $0.009 per input second, and the catalog lists one per-second rate; it does not list a separate rate per tier, so confirm in GET /v1/catalog.

The request fields

What the pricing text says

Set duration_seconds to the real length rounded up. An understated value gives you a reserve that does not match the clip, and a missing value reserves 5 seconds.

• Reserve is estimated from input seconds at $0.009 a second.
• If you omit duration_seconds, the reserve assumes 5 seconds.
• The maximum is 30 seconds per job.

A request that runs

Replace the URL with a public HTTPS clip you own. The call returns a job, so poll GET /v1/jobs/:id/status and read GET /v1/jobs/:id/result when it is ready. At $0.009 a second this 10-second clip reserves $0.09.

curl -X POST https://api.sume.com/v1/video-upscale-1.0/upscale \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: upscale-demo-001" \
  -d '{
    "video_url": "https://example.com/clip.mp4",
    "upscale_factor": 2,
    "duration_seconds": 10,
    "enhancement_tier": "standard"
  }'

Choosing the factor

A factor of 2 doubles width and height, so 1280 by 720 becomes 2560 by 1440, and a factor of 4 multiplies each side by four. Larger factors amplify the flaws in the source, so use the smallest factor that reaches your delivery size, and trial the tiers on one clip before you pick one for a batch.

Common mistakes

Writes need an Idempotency-Key, so reuse the same key when you retry the same upscale, and change it for a different clip. That way a network retry never bills twice for one clip. If a job fails, read the job envelope before you resubmit, because the error code says whether the input or the request was at fault.

• Sending model or endpoint fields, which the endpoint rejects because the server picks the provider model.
• Sending a clip longer than 30 seconds in one job. Split it first.