video-use

Transcribe source videos and prepare phrase-level transcripts for editorial decisions.
Propose a cut strategy, wait for user confirmation, then render preview and final edits.
Apply requested per-segment color grading, overlays, animations, and subtitles.
Review rendered cut boundaries for visual, audio, subtitle, and overlay issues.

Source video files in a videos directory.
User direction on content, target length or aspect, aesthetic direction, pacing, preserved or removed moments, animation or grading preferences, and subtitle needs.
An ElevenLabs API key for transcription.

An edit directory containing project notes, packed transcripts, an edit decision list, cached transcripts, animation assets, graded clips, subtitles, previews, final renders, downloads, and verification artifacts.
A preview render and, after confirmation, a final render.

The user must approve the plain-English editing strategy before cut execution.
Cuts must align to transcript word boundaries and include cut-edge padding.
Subtitles are applied after overlays.
Session outputs are written to the source videos edit directory rather than the video-use project directory.
Self-evaluation is capped at three passes; remaining issues are flagged to the user.
Inspect rendered output around every cut boundary for visual discontinuities, flashes, jumps, and waveform spikes.
Check that subtitles are not hidden by overlays and that overlay timing shows the intended frames.
Sample the first two seconds, last two seconds, and two to three midpoints for grade consistency, subtitle readability, and overall coherence.

AI 搜索