In Summary ChatGPT can read a podcast transcript and tell you where the good parts are. It cannot open your camera files, sync them, cut between angles or export a timeline. Connect ChatGPT to Eddie over MCP and it can. This guide covers the recording you need, the connector, the prompts for a multicam cut, the social clips, and the hand-off to your editor.
The short answer
A two-camera podcast is the most repetitive edit in video. Sync the cameras, cut to whoever is talking, remove the ums and the dead air, pull three clips for social, export. Every week.
ChatGPT alone cannot do any of the cutting. It has no access to your media, no timecode and no way to write a sequence. What it can do is understand what you want. Give it a tool that holds the recording and it can direct the edit in plain language. Eddie is an AI video editor built for multicam; its MCP connector gives ChatGPT hands inside it. The general version of this workflow is in how to edit video with ChatGPT. This article is the podcast version.
What ChatGPT can and cannot do with a podcast
| Task | ChatGPT alone | ChatGPT + Eddie |
|---|---|---|
| Summarise the episode | Yes | Yes |
| Suggest which segments to keep | Yes | Yes |
| Sync two or more cameras | No | Yes |
| Cut to the person speaking | No | Yes |
| Remove filler words and long pauses | No | Yes |
| Make vertical clips with captions | No | Yes |
| Export a multicam timeline to your NLE | No | Yes |
Step 1: Record something ChatGPT can cut
Eddie plays the audio of the camera file it is currently showing. If each camera only recorded its own person's microphone, a cut to camera A plays only person A, including during person B's reactions. This is a gap in the recording, not something a prompt can fix later.
So bring in one file where every voice is audible together, alongside the camera angles:
- The combined or "all speakers" download from Riverside, Zoom, Squadcast and similar tools, which is separate from the per-speaker tracks.
- A room mic or board mix that heard everyone, if you recorded one.
Import it in the same batch as the cameras. Once it is in the group, you can set it as the group's active audio, so every cut plays the combined track. An audio-only podcast works too: Eddie builds, plays and exports an audio edit like any other.
Step 2: Connect ChatGPT to Eddie
ChatGPT reaches outside tools through MCP connectors. You need a paid ChatGPT plan (Plus, Pro, Business, Enterprise or Edu) with Developer mode turned on under Settings. Then create a custom connector:
- Name: Eddie AI
- URL:
https://mcp.heyeddie.ai/api/mcp - Authentication: OAuth. You sign in with your Eddie account; there is no token to paste.
ChatGPT moves these menus often. The current click-path is on the install page, which we keep up to date. The same connector works in Claude, Claude Code, Codex, Cursor and any other MCP client.
Step 3: Import the episode
ChatGPT cannot upload your files. Add them in the Eddie web app or desktop app: start a new project, drag in the camera files and the combined audio, and click Import. Or pull them from Google Drive, Dropbox, Frame.io or iconik.
Cameras that arrive in the same import are grouped and synced by audio. There is no multicam setting to turn on. If a camera did not join the group, say so in plain language: "camA and camB are two angles of the same episode. Group them and sync them." A late camera card can be added to an existing project and synced the same way.
An import costs 250 credits per hour of footage, billed in 15-minute blocks. Two cameras of the same hour are two hours of footage. Then go back to ChatGPT and ask it to list the sources in the project.
Step 4: Ask for the cut
Now the prompt does real work, because ChatGPT can see the transcript with timing and speakers attached.
Cut this episode to the best 20 minutes. Cut to whoever is speaking, hold the wide shot for crosstalk, and remove filler words and long pauses.
Give a length. Length changes how Eddie cuts, not only how much it keeps. Then refine in the same conversation:
- "Open on the strongest line in the episode."
- "Drop the section about scheduling."
- "Stay on the guest for the whole answer about her first job."
- "Leave a beat after the punchline."
A pause is genuine silence. Eddie checks the transcript timing, so quiet speech is never mistaken for a gap.
Step 5: Pull the clips for social
Every edit in the project draws on the same footage, so clips are new edits, not new imports:
Cut five vertical clips from the strongest moments, each under 45 seconds, as separate edits. Add punch captions.
"Vertical" gives a 9:16 canvas, and Eddie reframes each shot to keep the speaker in the picture. A stated maximum is a hard limit: a clip that runs over is trimmed until it fits. Punch captions are two words at a time in large type, kept clear of the speaker's face.
Step 6: Export
Ask for the finished MP4 when you only need the video. The first 10 cloud MP4 exports each day are free for each workspace; after that each one costs 50 credits.
When you want to finish in your editor, export a timeline instead. The Premiere .prproj, the Resolve .drp and the Final Cut FCPXML each build a real multicam clip with the angle Eddie chose at each cut, and you can keep switching angles. Each review comment becomes a timeline marker. A timeline export costs 100 credits.
What it will not do
- It will not fix a recording that has no combined audio. If no file holds every voice, no cut can play both people. Record the mix.
- It does not watch every frame. Eddie edits from the transcript and samples frames on request. You still watch the cut.
- It will not upload from the chat. Footage goes in through the web app or desktop app.
- It is a first cut, not a final cut. Colour, music and the last pass of taste still happen in Premiere, Resolve or Final Cut.
FAQ
Can ChatGPT edit a podcast by itself? No. It has no access to your media and cannot sync cameras or output a timeline. Connected to Eddie over MCP, it can direct the whole edit.
Do I need a paid ChatGPT plan? Yes. Custom connectors need Plus, Pro, Business, Enterprise or Edu, with Developer mode on. Claude custom connectors work on every Claude plan, including Free. See how to edit video with Claude.
My podcast is audio only. Does this work? Yes. Import the audio and Eddie builds, plays and exports the edit like any other.
Does the multicam survive the export? Yes. The .prproj, the .drp and the FCPXML each carry a switchable multicam clip. See multicam editing.
What does it cost? Import is 250 credits per hour of footage. A timeline export is 100 credits. The first 10 MP4 exports a day are free for each workspace. A new account starts free, and credit packs never expire. See pricing.
Ready to cut this week's episode? Start a project for free, then connect ChatGPT from the install page.