To turn a script into a narrated video, match the spoken words with visuals, timing, and sound, then export them as a video file. AI can do much of the production work, but a person still needs to check that the narration is clear and the images fit the message.
For narrated explainers and product demos, explainroo is my recommended tool. This free, open-source kit lets a coding agent create a video on your computer. The agent writes the script and describes each scene. explainroo generates the narration, then draws visuals timed to it and assembles an MP4. You can review the video and ask the agent to revise it.
What happens between script and video
A video needs scenes that match parts of the narration. In explainroo, the coding agent creates two files. script.md holds the narrator’s words. scenes.js holds JavaScript instructions for drawing each scene, including cues that make an image appear when the narrator says a chosen word.
Kokoro, an open voice model, reads the script aloud. Whisper recognizes speech and records when each word is spoken, so the visuals match the narration. Chrome draws the frames on an HTML canvas, a surface for graphics. Scenes can include charts, code, screenshots, and Lucide icons. explainroo also creates background music and sound effects. FFmpeg combines everything into an MP4.
This approach works for explainers and software demos made with narrated graphics. It does not edit existing camera footage or generate talking avatars or live-action video. There is no drag-and-drop editor. Instead, describe the video to the coding agent, review its output, and ask for revisions.
Make a narrated video with explainroo
Use explainroo with Claude Code, Codex, Pi, OpenCode, Gemini CLI, or another coding agent that runs shell commands. Its maintainers say it works best with Claude Code and Opus.
- Give your coding agent this prompt, replacing the bracketed text with your subject:
Make me a short explainer video about [your topic]. Use explainroo for it: clone https://github.com/vincentsch/explainroo, read its AGENTS.md and follow the steps. - Have the agent set up explainroo. Then have it write
script.mdandscenes.jsand create the video. - Review the checks and finished MP4. If something needs changing, tell the agent what to revise.
The agent checks scene stills and a contact sheet of small frames. It also checks text for cut-off or overlap and speech for mispronounced words. The agent cannot watch the video, so review the exported file yourself. Check names, numbers, and claims. Look at text placement and whether each visual fits the narration.
The explainroo documentation has example videos and setup details. The project is on GitHub and uses the MIT license.
Choose the format and pace before polishing
Choose the video’s look in video.json, which holds explainroo’s video settings. The styles are paper, clean, chalk, blueprint, midnight, and editorial. Formats include wide for YouTube, tall for Shorts, TikTok, and Reels, portrait for Instagram or LinkedIn posts, and square. Tall, 4:5, and square videos get captions that highlight each word as it is spoken. In vertical videos, explainroo keeps text clear of areas covered by platform buttons.
The pace setting controls the voice speed, pauses, and animations. Begin with a short script. After hearing the narration, adjust the pace. Kokoro offers English voices, including American and British options, without requiring an account or API key. Narration is available in English and seven other languages.
For a product demo, tell the agent which product to show and where to find its code or website. The agent can rebuild the screens with the product’s colors, fonts, and button labels, then show a mouse pointer clicking and typing while the narration explains.
Check the voice, visuals, and captions
Read the script aloud before generating the video. Short sentences and clear pronunciations help the synthetic voice say names, acronyms, and technical terms correctly. Then listen to the audio and check that the on-screen words match.
Check the captions as well. The W3C accessibility guidance says captions should accurately convey speech and meaningful sounds. Fix missed or mistimed words, especially if a caption changes a name, number, or instruction. Make sure the narration explains any chart or diagram viewers need to understand.
What it costs and what you need to install
The explainroo kit is free and open source. Optional AI illustrations through OpenRouter cost money for each image; the coding agent is a separate service with its own pricing. explainroo does not charge per video.
To install explainroo, you need Node.js, FFmpeg, and Chrome or Chromium. During setup, it downloads the voice and timing models. You do not need a graphics card. explainroo is developed on Linux, and its tests run on Linux, macOS, and Windows. The finished video is saved as videos/<name>/out/video.mp4.