Agentcut lets users generate videos via AI prompts and code

From prompts to video scenes

Agentcut is a new tool that allows users to create videos by describing what they want in a chat interface. Instead of manually dragging clips onto a timeline, the system uses an AI agent to build the scenes, place motion graphics, and sync the music automatically. As agentcut.synteq.app reported, the tool is designed to turn a text description into a structured video project.

The process begins with a single command in a terminal after installing Node.js. Once the helper starts, the user can describe a video, and the agent begins writing real Remotion scenes. Remotion is a framework that allows developers to create videos using React, which means the video is essentially treated as code rather than a flat file.

This approach allows for a high level of customization because the agent isn't just generating a final video file. It is building the underlying components and laying out the timeline. This gives the user the ability to see the preview rebuild live as the agent works through the instructions provided in the chat.

Granular control and components

One of the primary features of Agentcut is the ability to maintain full layer control over the final product. While the AI does the initial heavy lifting, users can still drag clips, change properties in an inspector, or point at a specific frame to ask for a fix. This ensures that the user remains the director of the project rather than just a spectator of the AI's output.

The tool comes equipped with 49 motion components that are ready to drop into a project. These include specific elements like a Stat Counter that counts up to a value, a Chat Thread that builds bubble by bubble, and a Split-Flap Board similar to an airport departures sign. The agent knows these components by name and can deploy them based on the user's request.

Users can also apply styles to the entire project with a single word. This can restyle captions, graphics, camera moves, and textures across the whole video. For example, a user could take the same clip and edit and generate eight completely different videos just by switching the style, which allows for rapid iteration on different visual aesthetics.

Privacy and local execution

A significant technical distinction of Agentcut is that it operates entirely on the user's machine. The page has no backend, which means the agent, the project files, and any API keys stay on the local computer. This architecture ensures that the user maintains ownership of their data and the privacy of their creative assets while using powerful AI models.

The tool is compatible with both Claude Code and Codex, allowing users to choose which model they prefer to power their agent. Both models are given the same tools and skills to interact with the project files. This flexibility lets users switch between models in the chat depending on their preference or the specific needs of the video they are building.

Because the timeline is treated as a file and the video as code, the project is highly accessible to developers. Every scene is a React component in the user's workspace, meaning it can be opened in a standard code editor at any time. This transparency removes the 'black box' feel of many AI video generators, providing a clear path for those who want to dive into the underlying logic.

Open source and extensibility

Agentcut is released as an open-source project under the MIT license. This means that the studio, the helper tool, and all 49 components are available for anyone to fork, extend, and ship. This open approach encourages the community to build upon the existing framework to create new components or improve the agent's ability to interpret complex prompts.

The tool is designed to be a complete studio for prompt-to-video workflows. By providing the underlying code for the components, it allows for deeper integration into existing workflows. Users who are familiar with Node.js can set up the environment quickly to start generating content that is ready for export in wide and mobile formats.

Looking forward, the combination of AI agents and code-based video rendering suggests a shift in how content is produced. Instead of learning complex software to move objects on a screen, creators can focus on describing the desired outcome. The ability to then