Two hours of interviews.Six quotes on your timeline.
An agent works in a terminal card on your footage folder: pulls the audio with ffmpeg, transcribes every interview locally with Whisper, and picks the quotes that answer the producer’s brief. You change the list by voice, it writes an EDL and FCPXML your editor imports, and cuts the selects without touching the originals.
You swap one quote by dictation. Allow the cut when it asks.
Three interviews for a hospital documentary: the agent transcribes them, picks quotes against the brief, takes your dictated change and exports an EDL and selects for the edit.
Your footage stays put
Everything runs in a terminal card on your own disk: ffmpeg, the Whisper CLI and a small script the agent writes for the edit list.
Quotes against the brief
The brief is a note on the canvas. Each pick comes with the clip, the timecodes and the part of the brief it answers.
Change it by voice
Dictation turns “swap Mark’s last quote” into text in the agent card — no typing between two monitors.
Import, don’t retype
An EDL and FCPXML go straight into your NLE; the selects are stream copies, and the originals are never rewritten.
From the card reader to the first assembly
One agent, four cards.
Imports into the editor you already use
More in Video editing
Other live scenarios from the same shelf — and a few close relatives.
Batch deliverables
Video editingOne master rendered to every platform: vertical crops with captions, loudness to spec, nothing overwritten without asking.
AgentTerminalNoteTo-do listReferenceSubtitles and translations
Video editingAn SRT from the transcript, two translations, and a reviewer on another AI checking timing and terms.
AgentTerminalNoteTo-do listData questions
Beyond codeHypotheses on a list, SQL on a read-only replica, an answer with its queries.
AgentBrowserTerminalNoteTo-do listWorkflows
Beyond codeDaily prompts, web watches and Telegram messages that set agents to work.
AgentBrowserTerminalNoteTo-do listTelegramStandup
Stop scrubbing for the one good sentence
The agent asks before it writes a single file into your project.
macOS and Linux are coming later.
The first-run wizard installs an agent CLI, Git and the dictation model for you — or skips each step.