Read an AI video's transcript before posting: Sume video inspect

Spoken claims in an AI video are easy to miss. Use Sume video inspect with transcribe to read what was said before you post on TikTok or LinkedIn.

4 min readSume
All posts

Send the finished file to Sume's video inspect with transcribe: true and read the text before you post. A model with native audio can say things you did not write, and platforms are tightening checks on AI content, so reading the transcript is a cheap review step at $0.01 per audio minute for the speech-to-text part.

The route and the rates are from the Sume video inspect guide. The platform context is TikTok's AI transparency post and LinkedIn's feed quality write-up, both read on 2026-10-07.

What the call returns

One request can do three jobs.

Sume video inspect (read 2026-10-07)
PartWhat you getCost
ProbeDuration, size, probe.has_audioNot stated as a separate fee in the guide
StillsUp to 24 frames as durable media.sume.com imagesNot stated as a separate fee in the guide
TranscriptText with word timings from Sume STT 1.0, with transcribe: true$0.01 per audio minute

How to use it

The source must be a media.sume.com artifact or asset, up to 1800 seconds. Two errors are worth knowing: video_inspect_transcribe_required if you send transcript-only fields without transcribe: true, and inspect_source_has_no_audio if you ask for a transcript of a silent clip. If you are not sure whether a clip has sound, run a probe-only call first with frames: false and read probe.has_audio.

What to look for in the text

Read it as a reviewer, not a viewer.

  • Numbers, prices, dates and names that you did not supply.
  • Health, money or political statements, which TikTok says it is testing AI-spam detection on.
  • Phrases that repeat across videos in a batch, which LinkedIn's classifiers may treat as low-quality.
  • Claims about the product that your legal or brand team has not approved.

Then fix and recheck

If a line is wrong, regenerate with a tighter prompt, or set generate_audio: false and add a voice track you wrote. Burned captions with cues from video captions put your approved words on screen. For the niche-specific checklist see the TikTok spam test post, and for a webhook-driven review queue see the approval gate.

Sources

Related posts

More in Media tools

All Media tools posts

Written by Sume