Soften skin in video by API with smartblur (whole-frame)

smartblur is in the Blur / sharpen / noise group of Video filter, and FFmpeg describes it as blurring the input video without impacting the outlines. That makes it the closest allowlisted filter to a skin-smoothing pass, with one important limit: Sume's video filter has no face detection or mask, so it softens the whole frame, not just skin. Use it for close beauty shots where that is acceptable.

Which smartblur options set the strength?

How do I keep detail while softening?

Use a small radius, a strength below 1 and a positive threshold so edges such as eyes and lips are left alone. Check the graph, encode a few seconds, and compare a still of the source with a still of the result before applying it to the whole clip.

curl -X POST https://api.sume.com/v1/video-filter/check \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "video_url": "https://media.sume.com/artifacts/artf_demo/beauty.mp4",
    "filtergraph": "smartblur=lr=2:ls=0.7:lt=5"
  }'

Can Sume smooth only a face?

Not with the video filter. It takes one clip and a filter program, with crop and dim ops and an allowlisted graph, and none of them track a subject. If the face stays in one place for the whole clip, a graph can split, crop that region, soften it and overlay it back; see the magnifier inset graph for the same pattern with fixed coordinates. For a moving subject, there is no tracking option.

Should I disclose retouched footage?

Platform rules on retouching and AI labels vary and change. Read the destination platform's own policy; Sume's docs do not state a rule for retouched footage, and a pixel filter does not add any marking by itself.

What does a filter job cost and where are the limits?

The docs list $0.02 per encode job, with a source of at most 300 seconds, a filtergraph of at most 2048 characters and 32 named filters. The /v1/video-filter/check route runs the same allowlist and source checks for free and returns valid and next_action; a graph that passes can still fail on the worker (bad expression, memory, time), which comes back as a structured job error.