Reference video analysis on Sume: inspect, ingest or video-analyses?
For a reference video, use Sume video inspect for stills and probe, reference ingest for shots and OCR, and note that video-analyses returns 410 on dest.
For reference video analysis on Sume, pick by what you need. Use video inspect for a probe, a few stills and an optional transcript. Use reference ingest for shots, on-screen text, audio loudness and cut boundaries from a clip of 300 seconds or less. Do not build on the older POST /v1/video-analyses: it returns 410 video_analysis_retired on dest and stays on prod only until its retirement lands. Sources: Video inspect, Reference ingest and Video analyses, read 2026-10-06.
How do they compare?
Three endpoints, three jobs.
| Endpoint | Returns | Limit | Price |
|---|---|---|---|
/v1/video-inspect | Probe, 8 stills by default, optional STT | Source up to 1800 s | Modal compute; STT $0.01 per audio minute |
/v1/reference-ingest | Shots, OCR text tracks, audio, boundaries, overview strip | Source up to 300 s | Modal compute |
/v1/video-analyses | Legacy analysis | Retiring | $0.30 on prod; 410 on dest |
What gates reference ingest?
The route sits behind SUME_COM_REFERENCE_INGEST_ENABLED, dest first with a production opt-in, so check that it is on for your environment before you call it. Refusals you can plan for are source_too_long_for_reference_ingest and reference_ingest_stt_required.
Sources
Related posts
More in Media tools
- Same music under every Shorts episode: Timeline soundtrack settings
Sequential playback means viewers hear episodes back to back. Reuse one music file at one gain and ducking level in every Sume Timeline render of the season.
- Shorts episode title card: a 2-second still, then the clip
Open each Shorts episode on a held title still with Timeline 1.0: slot math, the 0.5 s coverage rule, what stills ignore, and one render body you can copy.
- Shorts spec sheet against Sume Timeline output: what is confirmed
YouTube lists 3 minutes, 1080p and a 100-character title for Shorts. Sume Timeline renders 1080x1920 by default. A table of what is confirmed and what to check.
- Silent AI clip fails caption_no_speech: burn text with cues instead
A clip with no speech fails the Sume caption job with caption_no_speech. Send cues with text, start and end to burn authored overlay text for a Short or ad.
Written by Sume