Claude can now edit video: from raw footage to a finished cut

When people think of AI and video, they usually think of models that generate footage from scratch. In recent weeks something else has been getting attention: Claude can take existing footage and materials, edit them, animate them and deliver the result as a video file. And it does so with your material, without inventing new footage.

1. Claude Motion: animations built from code

On 8 October 2026 Anthropic announced Motion in beta. Motion turns a report, a chart or a product walkthrough into a short animation, for example a quarterly report into a 30-second explainer, or a chart from a deck into a moving scene.

• It doesn't use a video generation model. According to Anthropic, Motion produces no generated footage and no AI-generated people; it animates your own text, charts, shapes and images.
• Every animation is built as code, so after the first version you can change any word, number or timing.
• It downloads as MP4, and editing can continue in tools such as Adobe, Descript or Runway.
• For now it is offered in beta on Team and Enterprise plans.

2. A 10-minute vlog from raw footage

The example that made the most noise came on 1 November. Entrepreneur Pieter Levels shared that his vlog that day was edited entirely by Claude Code. What he did was simple: he put 14 camera clips into a folder, opened Claude Code there and asked for a 10-minute vlog for X and YouTube. He didn't say which methods to use.

According to Levels, Claude Code set up the process itself:

• It transcribed the speech with a Whisper model running on the laptop.
• Instead of watching every minute, it scanned the footage through contact sheets of frames sampled every few seconds.
• It used voice recognition to separate Levels' voice from other speakers.
• It wrote the cuts as a list (clip, start, end) and rendered the video with a script.
• It prepared thumbnail and caption images using face detection.

The result: about 58 minutes of footage across 14 clips came down to roughly 10 minutes. Levels says he made only a few small adjustments.

We're trying it in our own work

We've been producing Dada Studio's Instagram reels with a similar approach for a while. We build scenes as code with our design files and project images, render the frames and turn them into video. We automatically find the pauses between sentences in the voiceover and fit each scene's length to the audio. So a revision, such as holding a scene two seconds longer or speeding up the whole video, takes minutes instead of hours in an editing program.

There are limits too

• Anthropic hasn't released an editing program. In the vlog example, Claude Code wrote and ran scripts that call open-source tools. The work happens in code, not on a timeline.
• The results are self-reported. Levels' raw footage isn't public, and some of the tools used only run on Apple silicon.
• Creative judgement is still human. Deciding which moment carries the story, the brand's tone and the rhythm of the music, and doing the final check, remains the job of an editor or designer.

What does it mean for brands and agencies?

Faster and cheaper. For recurring video work such as product teasers, event recaps or interview clips, a first cut can now be ready in minutes.

Brand consistency is easier to keep. An animation built in code can repeat the same colours, typefaces and transitions in every video. That means turning a well-defined brand identity into a system on video too.

The real difference is the input. AI edits the material you have; providing the material itself, well-shot footage, a strong story and a clear brand language, is still our job.

If you're planning social videos or product content for your brand, take a look at our social media and post-production services or request a quote.

Source: Unite.AI, October 2026

Source: The Decoder, October 2026

Source: Pieter Levels, X, 1 November 2026

Source: CellCog, November 2026