Case study

Cut Video Gaps at a TV studio: a 35-minute recording cut to 24 minutes on the Mac

Cut Video Gaps is a macOS app we designed and built in-house. Rede Raseck TV, a digital TV production studio in Brazil, is its first customer case study.

  • Electron
  • React
  • TypeScript
  • FFmpeg
  • faster-whisper
  • Vite
  • Tailwind
  • Python
  • macOS
Cut Video Gaps result screen: a 35m 22s recording cut to 23m 56s, 11m 26s saved, exported as 315 clips
  • 11:26removed from a 35:22 recording
  • 315editable clips in Final Cut Pro
  • ~6 hper episode, down from ~14 h (team's estimate)
  • 4 daysfrom a 3-star review to the fix

The problem: about 14 hours of editing per episode

Rede Raseck TV is a digital TV production studio in Brazil. Its shows are shot in 4K against a green screen, placed on virtual sets with sponsor graphics, and edited in Final Cut Pro. The recordings are long, unscripted talks in Portuguese.

On 20 October 2026 the team edited a 35-minute recording by hand. They estimate one episode took about 14 hours of editing.

Desktop app development: what Cut Video Gaps does

  • Fast Cut: silence detection with FFmpeg from a loudness envelope, for a quick first pass
  • AI Auto-Cut: the recording is transcribed on the Mac, then filler words, repeated words and possible retakes are marked for review
  • Exports to Final Cut Pro (FCPXML), Premiere Pro (XML) and After Effects that reference the original media: frame-accurate, source timecode kept
  • Private by design: footage and transcripts never leave the user's Mac

From a 3-star review to a fix in four days

  1. The team edits a 35-minute recording by hand in Final Cut Pro.

  2. First test: Fast Cut on v1.1.0, a 35-minute Apple ProRes 422 file on an M1 MacBook Pro. Rating 3/5, and the whole review was “slow.” The video export was the slow part.

  3. We replied personally and asked what was slow: which mode, which file, which Mac. v1.2.0 shipped with much faster exports for videos with many cuts.

  4. Second test, AI Auto-Cut on v1.2.0: 35:22 cut to 23:56 and exported to Final Cut Pro as 315 clips in under a minute.

  5. The team nicknames the app “Navalinha” and approves this case study in writing.

31-second motion-graphics summary of the Rede Raseck TV case study, no voice-over

Engineering: Electron, FFmpeg and on-device AI

  • Electron, React and TypeScript

    A macOS desktop app for Apple Silicon built with Electron, React, TypeScript, Vite and Tailwind. Signed, notarized and auto-updating, so a fix reaches every user without a reinstall.

  • FFmpeg as the media engine

    Silence is found from a loudness envelope. Video exports re-encode at the exact frame with hardware encoders; audio-only exports are lossless.

  • On-device AI with faster-whisper

    Transcription runs in Python, packaged into the app with PyInstaller. Tested in English, Hindi and Punjabi (bad-take detection also in Urdu); this customer's recording was in Portuguese.

  • Export rework in v1.2.0

    Frames are routed to their cut group before the exact trim. In our test a 17-minute interview with 355 cuts exported in 28 seconds instead of 7 minutes on an M1 Pro, same frames and audio: about 15x faster.

  • Editor hand-off without a re-encode

    The Final Cut Pro export points at the original ProRes file instead of rendering a new one, so a long cut opens in the editor quickly and every cut stays editable.

  • No upload step

    Footage and transcripts stay on the Mac, so a large ProRes file is processed where it already sits and never goes to a third party.

Dr. Raseck of Rede Raseck TV
“Of all the tools I have tested in this category, Cut Video Gaps is the best I have ever used and the only one I have truly approved. Its efficiency, precision, and integration with my editing workflow have genuinely impressed me.”
Dr. Raseck, Rede Raseck TV

The result on a real recording

  • 35:22 recording cut to 23:56, with 11:26 removed
  • 2,969 Portuguese words transcribed on the Mac; 49 filler words, 11 repeated words and 53 possible retakes marked for review
  • AI Auto-Cut took about 5 minutes on an M1 MacBook Pro
  • Exported to Final Cut Pro as 315 clips in under a minute, pointing at the original ProRes file, every cut editable
  • The team estimates an episode now takes about 6 hours instead of 14 (their estimate, not a timed comparison)
  • About ten days after the 3-star review, the team nicknamed the app “Navalinha”, Portuguese for “little razor”, for how cleanly it cuts
  • Next: the customer's log showed AI Auto-Cut spends most of its time reading audio from large ProRes files, and we are working on making that faster

Need something similar?

Rede Raseck TV's hand edit in Final Cut Pro: the finished show on a virtual set
Before: the team's hand edit from 20 October 2026, the finished show on a virtual set.
The Cut Video Gaps cut opened in Final Cut Pro in the Rede Raseck TV library, as a timeline of separate clips
After: the Cut Video Gaps cut opened in Final Cut Pro, in the Rede Raseck TV library, with every cut editable.

Published with Rede Raseck TV's written approval. They use the free public beta and received 5 bonus processing hours after sharing feedback.

What we learned, written up

More case studies