cutproofstudio
LOCAL-FIRST
Source-linked short-form editing

Keep the context. Cut the rest.

Find useful moments, keep their source intact, and hand off clips with receipts.

No account. No API bill.Your files stay in this browser.
Nothing is uploaded or published.
WAITING FOR SOURCE
Your source, on the record.

Load the demo or attach a video
to preview exact source ranges.

Nothing has been uploaded.00:00
Source first. Suggestions second.
Import captions and select Find source-linked clips.
BOUNDARY LAB / ACTUAL SOURCE WORDS

What does this cut leave out?

Compare the selected excerpt with a wider source range. Added cues stay verbatim. Applying a suggestion never marks it as reviewed.

From source to a real handoff.

Import. Open a timestamped SRT, VTT, or JSON transcript. Attach the video from the same recording. No transcription model runs here.

Select. The local ranker uses corpus TF-IDF, sentence structure, and diversity. Selections contain whole, consecutive source cues. English lexical checks flag possible missing context.

Review. Inspect the earlier and later cues in Source trace. Change boundaries when needed. Review states start as pending and reset when boundaries or the attached video change. Boundary Lab shows the omitted words and proposes an expanded range. Saved projects restore with all reviews pending until the original video is reattached and checked.

Export. The ZIP includes a manifest, source cues, captions, verbatim publication drafts, and a Python MP4 renderer. In supported browsers, Render WebM records a portrait version locally in real time. Keep the tab visible during recording.

No paid service, pretrained model, cloud upload, auto-publication, or claim of guaranteed engagement is included. Source alignment is not proof of truth.

Paste timestamped captions

SRT, VTT, a JSON array of {start, end, text} segments in seconds, or {start_ms, end_ms, text} cues.