When you upload, ClipAd analyzes the whole video and proposes a set of clean segments — each one a coherent scene with a natural start and end. You never touch a timeline.
What the engine detects
The AI works from what is actually said. Your audio is transcribed with word-level timings, and a language model reads that transcript to find the moments that stand on their own — a complete point, a story, a question and its answer. Clips are trimmed to real sentence boundaries, so one never starts or ends mid-thought, the way a good editor would carve up a long video.
Because it reads the transcript rather than watching for visual cuts, detection needs speech. If it can't complete — say the transcription service is briefly rate-limited — the video is marked Detection failed with the reason and you can retry. You will never be handed arbitrary, unranked clips in place of a real result.
Why each clip gets a score
Every proposed segment is scored for how well it fits advertising — its potential as a high-retention, ad-friendly moment. A strong hook or a clear product demo scores higher than dead air. The score is guidance, not a verdict: it helps you decide what's worth keeping. See Understanding segment scores.
You stay in control
Detection is a proposal, not a decision. Nothing is monetized until you review the segments and approve them. You approve the clips worth keeping and reject the weak ones — the AI does the tedious first pass, and the editorial calls are yours.