Claude + video
Can Claude edit my videos?
Partly. Claude can plan an edit and run the tools that cut it, but it can't watch your video. Anthropic's own docs list pictures as what Claude can see: JPEG, PNG, GIF and WebP, not video files. So every "Claude edits my videos" setup first turns your footage into a transcript and a few still frames. It also takes a lot of setup. That works well for talking heads. It struggles with anything that depends on the picture or the music. OttoEdit is plug and play, and it watches the video itself.
Where the "Claude edits my videos" idea comes from
- Claude Code plus ffmpeg. A speech-to-text tool writes a transcript, Claude writes a list of cuts, and ffmpeg, a free video tool, renders them. Good for cutting silences and retakes out of talking videos. You run it in a terminal.
- Remotion and HyperFrames. Claude writes code that renders titles, animated captions and motion graphics. That's making new video from code, not cutting your footage.
- Claude inside an editor. DaVinci Resolve Studio 21.1 lets Claude run steps in Resolve, and community projects do the same for Resolve and Premiere. More on how that works.
- Skills and plugins on GitHub that bundle these steps together. Most need installing and setting up.
What Claude can't do on its own
- Watch your footage. It works from the transcript and whatever frames a tool pulls out. Sending it one frame a second of 10 minutes of 1080p footage costs about $6 in Claude usage, by one October 2026 test, so most setups send far fewer.
- Hear your song. Speech can be turned into text for Claude. A beat, a drum hit or a guitar solo can't. Cutting to music needs other tools.
- Judge the picture. The same test found it suits speech-led videos, and recommends keeping multicam, colour, b-roll picked by look, and sports or action footage in a real editor.
- Hand you a finished edit. Expect a draft to check. Speech-to-text timings can clip the start or end of a word by about half a second, which you hear as a chopped cut.
The setup it takes
A typical "Claude edits my videos" setup looks like this:
- A paid Claude plan, then Claude Code installed and run from a terminal.
- ffmpeg installed, plus a speech-to-text tool like Whisper, which needs Python.
- A skill or script from GitHub to tie them together, and the right versions of each.
- Fixing what breaks. The October 2026 test above hit a speech-to-text crash that needed an older version of one library, and an ffmpeg build missing the captions feature. Installing the fix broke another tool until that was updated too.
The Resolve 21.1 route is lighter, but you still pay for Claude or ChatGPT, install its desktop app, and connect it under File > Setup AI Assistants.
OttoEdit is plug and play. Install it, sign in, and flip one switch in Resolve's preferences, which Otto shows you. Then open your project and talk. No terminal, no code, no keys, and the AI is included.
How OttoEdit is different
- It actually watches. Otto sends your clips, with their sound, to a vision model that watches the video itself, not just stills. A reasoning model then plans the edit from what was seen and heard. Speech-to-text runs on your own computer.
- It hears the music. The beat, the drums, the bass and the vocal, so it can cut on them.
- It checks its own work. Before a new cut lands, Otto watches its draft back with the song, and redoes it if it isn't good enough.
- A real Resolve timeline. Not a rendered file. The edit sits in your DaVinci Resolve project, ready for your notes or your own hands.
- Plug and play. No terminal, no ffmpeg, no keys to paste. Open Otto next to Resolve and talk to it. The AI is included, and it's free right now.
When plain Claude is enough
If you're happy in a terminal and your videos are mostly people talking, a Claude Code setup can save real time on silences and retakes. For code-made motion graphics, Remotion with Claude is great. If you edit in DaVinci Resolve and want the whole edit made from footage it has actually watched, that's what Otto is for.
Questions
Can Claude watch a video?
Not directly. Anthropic's docs list images as what Claude can see. Tools that "let Claude watch" a video pull out still frames and a transcript first.
Can Claude edit videos in DaVinci Resolve?
Yes, with Resolve Studio 21.1 or newer, after you connect it under File > Setup AI Assistants. It runs steps you ask for. Blackmagic doesn't say it watches your footage.
Is Claude Code good for video editing?
For talking-head videos, as a draft you check, yes. For music videos, b-roll, colour and anything that depends on the picture, it needs other tools.
Does OttoEdit need any setup?
Barely. Install it, sign in, and turn on one switch in Resolve's preferences. Otto shows you where. No terminal, no code and no API keys.
How does OttoEdit see my footage?
A vision model watches your clips with their sound, and a reasoning model plans the edit from that, so the plan is built on footage that was actually seen and heard.
Try it on your own footage. It's free right now.
Get early access