If you need a product demo but do not want to build every scene by hand, a coding agent can turn a short brief into a narrated video. A coding agent is an AI assistant that runs commands and works with files. It can write a script or set up a video project.
For narrated explainers and product demos, explainroo is the best fit: this free, open-source kit lets a coding agent create videos on your computer. The agent writes the narration and describes the scenes; explainroo renders them into an MP4. Its product demos rebuild screens from your product’s code or website, so review the result against the current product before publishing.
What a coding agent makes possible
AI product video generators include several kinds of software. Some create new video imagery from prompts; others turn product information or screen recordings into a finished video. In a coding-agent workflow, the assistant writes and revises video files, then runs software to draw scenes, add narration, and assemble the video.
This approach works for explainers and product demos made from scenes, text, and graphics. The video does not necessarily show a live recording of the product. With explainroo, the agent rebuilds product screens based on the code or website you provide, then adds a pointer that clicks through screens and types into fields. This can show a workflow clearly, but you should check labels, colors, and actions against the actual product.
Make a product demo with explainroo
Start by opening a coding agent that can run shell commands. explainroo works with Claude Code, Codex, Pi, OpenCode, Gemini CLI, and other agents; it works best with Claude Code. Paste this prompt, replacing the bracketed topic:
Make me a short explainer video about [your topic]. Use explainroo for it: set it up, read its AGENTS.md, and follow the steps.For a product demo, tell the agent the product name and where to find its code or website. The agent sets up explainroo, creates the video, checks it, and gives you an MP4. See the explainroo documentation and example videos for its workflow and examples.
The agent writes two files. script.md contains the spoken words. scenes.js describes what each scene shows and can link a drawing to a chosen word in the script. You can ask the agent to revise either file when a line or scene needs work.
explainroo uses Kokoro, an open voice model, for English narration, with natural American and British English voices. Whisper listens to the recording and marks when each word is spoken, so scene changes can align with narration. Chrome draws the scenes on an HTML canvas; explainroo then combines the video and sound into an MP4. You can include code, charts, and screenshots. Its five visual styles are controlled by a setting in video.json.
Shape the video and check the result
Choose a format that fits where you plan to post the video. explainroo supports wide video for YouTube and tall video for Shorts, TikTok, and Reels. It also supports portrait video for Instagram and LinkedIn posts, plus square video. Captions light up word by word in tall, portrait, and square formats. A pace setting adjusts narration, pauses, and animation speed.
Before publishing, inspect the product screens and listen to the narration. explainroo gives the agent scene stills and a contact sheet. It also checks for clipped or overlapping text and mispronounced words. These checks catch some problems, but the agent cannot watch the finished video. Review it yourself, especially for product claims and customer data. In the United States, the FTC’s advertising guidance says advertising claims must be truthful and supported by evidence.
Cost and limits
explainroo is open source under the MIT license. It runs on your computer and does not require a graphics card. Its voice and timing models download during first-time setup. The video-generation steps cost nothing per video; optional AI illustrations through OpenRouter are paid per image, and the coding agent is a separate service with its own pricing.
The kit makes narrated explainers and product demos. It does not edit existing camera footage, create talking avatars, or generate live-action video. There is no drag-and-drop editor; you describe changes to the coding agent instead. To install it yourself, you need recent Node.js and ffmpeg, plus either Chrome or Chromium. The project repository lists the setup commands.
