A set of five small tools that automate the boring parts of corporate video editing: transcribing interviews, finding the good soundbites, archiving finished projects, turning messy client feedback into a checklist, and one-command encodes for stage playback.
You do not need to be a programmer to use these. Every tool can be run by dragging a folder or file onto a script, or by double-clicking. Each folder has its own README with step-by-step instructions and a "when it breaks" section.
- Install Python 3.11 or newer from https://www.python.org/downloads/. On Windows, check the box that says "Add Python to PATH" during install.
- Open the tool folder you want to use and follow its README's setup section
(usually a single
pip install -r requirements.txt— the README shows exactly what to type or double-click).
That's it. No accounts, no subscriptions, nothing goes to the cloud except plain text in the two tools that explicitly say so.
| Folder | What it does | Footage leaves your machine? |
|---|---|---|
transcribe/ |
Drop a folder of interview clips → get timecoded transcripts (.txt, .srt, .json) for every clip. Runs entirely on your computer. | No — 100% local |
story-assist/ |
Transcripts + the client's story brief in → recommended soundbites with timecodes and a suggested story order out. | Only transcript text (never video/audio) |
archiver/ |
Verified, resumable copies of finished projects to your external drives, plus a searchable index of which drive every project is on. | No — 100% local |
feedback-parser/ |
Paste messy client emails / Frame.io comments → clean, timecoded revision checklist grouped by category. | Only the feedback text you paste |
encode-presets/ |
One-command H.264 encodes for delivery, including odd stage-wall resolutions (pad/crop/stretch). | No — 100% local |
RAW footage never goes to the cloud. The two AI-assisted tools (story-assist
and feedback-parser) send only text — transcripts and feedback notes — to
the Anthropic API. Video and audio files are never uploaded by anything in this
repo.
Your most-hated task is already solved by software you have:
- Premiere Essential Sound panel: select all your dialogue clips → Essential Sound → tag as Dialogue → Loudness → Auto-Match. This levels every clip to a broadcast-standard loudness in one click.
- Enhance Speech (Premiere 2024+): in Essential Sound under Dialogue, toggle Enhance Speech for noisy or roomy recordings. Use the strength slider sparingly (30–50% usually sounds natural).
- For genuinely rough recordings (phone audio, echoey conference rooms), run the file through Adobe Podcast Enhance (free with your Adobe login at https://podcast.adobe.com/enhance) and cut the cleaned file in instead.
- Final check: keep your master mix around -23 LUFS for broadcast or -16 LUFS for web/ballroom playback (Premiere's Loudness Radar or the Essential Sound auto-match handles this for you).
You'll receive new versions as a zip file with a dated changelog
(CHANGELOG.md). To update: delete the old folder, unzip the new one, done.
Your settings files (.env, config.js, archive index) live outside the
zip's replaced files or are listed in the changelog if you need to carry
anything over.