What You Can Ask

You don't need to know tool names. Describe what you want the way you would to an editor, and the assistant picks TimeBolt's tools. Below are requests that work well, grouped by task, with a short note on what TimeBolt does for each.

Tighten a talking-head video

Open /Users/me/Movies/episode-12.mp4 in TimeBolt. Remove the dead air, the filler words, false starts and retakes, then render it.

The assistant opens the file (which runs silence detection), transcribes it on your computer, detects filler words, false starts and retakes, cuts those words from the timeline, then renders and reports back when the file is ready.

More you can ask:

• "The cuts feel too tight. Run silence detection again with 0.2 seconds of padding."
• "What filter level would TimeBolt pick automatically for this file? Use it."
• "Only remove silences longer than half a second."
• "Fast-forward the silences at 2x instead of removing them."
• "Show me the filler words you found before cutting anything."
• "Add 'basically' and 'kind of' to my filler word list."

Make specific edits

Remove everything between 2:10 and 2:45, and keep only the cuts I starred.

The assistant reads the current timeline and changes only the cuts you mention, so the rest of your edit stays as it is.

• "Cut the part where I say 'let me check my notes'."
• "Split the timeline at 4:30 and fast-forward everything after it."
• "Zoom in 1.5x, top-left, on every cut between 1:00 and 3:00."
• "Mark the cuts where I talk about pricing."
• "Merge consecutive cuts that have the same settings."
• "The source was trimmed by 3 seconds at the start. Shift all cuts 3 seconds earlier."
• "Play the preview from 5:20 so I can check that cut."
• "Undo that." (Every timeline change can be undone, also with Cmd/Ctrl + Z in TimeBolt.)

You can also save and reload your work:

• "Save the timeline as JSON next to the video."
• "Load the timeline from /Users/me/Movies/episode-12.mp4.json."

Chapters

Read the transcription and add a chapter at the start of each new topic, with a short title. Then write a YouTube chapters file.

• "Render one clip per chapter."
• "Give me the chapter timestamps for a 1.25x Turbo Mode render."

Captions

Transcribe this video, change every "Time Bolt" to "TimeBolt", then export SRT captions.

TimeBolt transcribes on your computer (the first time, it downloads the speech model) and the SRT follows your edited timeline, not the original file.

• "Here is my transcript as JSON: … Use it instead of transcribing." (Word-level, as [{ "start": 0.0, "end": 0.4, "word": "Hello" }]. See the Tool Reference.)
• "Save the transcription as JSON."
• "Render the video with the captions burned in."

Export to your video editor

Export this timeline as XML for Premiere Pro.

Files are saved next to your video unless you name a location. TimeBolt never overwrites a file: if the name is taken, it adds a number. For an audio-only file, tell the assistant the frame rate your editor uses (for example "…at 29.97 fps"), since XML and EDL files need one.

Render

Render this as an MP4 at 1.2x speed into /Users/me/Exports.

• "Quick render as a WAV file."
• "Add this to the render queue, then add only the starred cuts as a second render, then start the queue."
• "Add background music from /Users/me/Music/bed.mp3 at 20% volume, then render."
• "Export every kept cut as its own file into /Users/me/Clips."
• "How far along is the render?" / "Cancel the render."

Renders start straight away and run in TimeBolt. The assistant checks progress and tells you when the file is ready.

Process many files at once

Batch process every video in /Users/me/Course/Module 3 with my current silence detection settings.

The assistant opens each file in turn, queues a render for each, and starts the queue. This replaces the project open in the main editor.

Multi-Track Editor: podcasts and multi-camera

Open cam-a.mp4, cam-b.mp4 and mic.wav in the Multi-Track Editor, sync them, remove the silences, and switch to whichever camera is speaking. Then export an FCPXML.

The assistant adds one track per file, synchronizes them by their audio, runs silence detection on each track, and keeps one camera angle at a time (the top-most kept track is what the video shows).

• "Use the mic track to decide the cuts and apply them to every track."
• "Keep everyone in frame whenever anyone is speaking."
• "Start track 2 at 1.5 seconds." / "Swap tracks 0 and 1." / "Mute the camera audio."
• "Transcribe all tracks and cut to the person who is speaking."
• "Render the multitrack edit to /Users/me/Exports/episode.mp4."

Multi-Track Editor cuts can only be keep or remove (no fast-forward or zoom). Removing a track cannot be undone.

Record your screen

Record my main screen with my camera and microphone, square crop, without system audio. Start now.

Later: "Stop the recording and save it to /Users/me/Movies/take-1.mp4." TimeBolt counts down three seconds before recording, then hides its windows. When saving finishes, the recording opens in the main editor, ready to edit.

On Mac, give TimeBolt screen recording permission in System Settings > Privacy & Security before your first recording.

Fix problem files

Can TimeBolt open /Users/me/Downloads/zoom-call.mp4? Check it for variable frame rate and fix it if needed.

• "Convert it to a constant frame rate and open the result in TimeBolt."
• "Re-encode this file, my editor can't read it."

See Issues with Variable Frame Rate Files for why this matters.

Settings and windows

• "Turn on CUDA GPU rendering."
• "Set the FCPXML export version to 1.11."
• "What are my current silence detection settings?"
• "Open the Stitch window." / "Close the settings window."
• "What is TimeBolt doing right now?"