Fix washed-out, flat video: normalize or histeq filter API

Send the clip to POST /v1/video-filter with normalize or histeq in the filtergraph; both are allowlisted tone filters. normalize stretches the tonal range automatically, histeq equalizes the histogram, and colorlevels lets you set black and white points yourself. The encode is listed at $0.02 per job.

Context: Runway's changelog, read 2026-10-01, says its Ruby model in Tool Mode (Aug 31, 2026) lets you bring in outside videos and color-grade them. That is a model-driven grade. This post covers the plain filter route, which has no model and no prompt.

Which filter should I try first?

Start with the automatic ones, then fall back to manual levels if the result is too strong. These are ffmpeg filters, so their option names are ffmpeg's.

How do I run it?

Check first. It validates the allowlist and the source without creating a job or reserving credits.

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/flat.mp4",
    "filtergraph": "normalize"
  }'

What are the limits?

The source must be 300 seconds or shorter and on media.sume.com. The graph can hold up to 2048 characters and 32 named filters. An auto-level pass works on each frame or the clip's range as the filter defines it, so a clip that mixes bright and dark scenes can shift between shots; for those, cut with video trim and fix each shot separately.

Is this HDR conversion?

No. The docs describe the output as a new MP4 with source geometry, frame rate and audio; they do not describe HDR or a color-space change. For a broader grade, see color-correct a video.