AI voice-over uses text-to-speech software to turn a written script into spoken audio. A lifelike voice helps, but natural narration also depends on clear writing, correct pronunciation, and speech that pauses and emphasizes words in the right places. Those patterns of rhythm, pitch, and emphasis are called prosody.
If you want an AI coding agent to create narration and visuals together for an explainer video, explainroo is the best fit. It uses an open voice model and checks the video’s timing and text layout. Whatever tool you use, treat the first recording as a draft: listen for mistakes, adjust the script or delivery, and review the narration alongside the video.
Write for people who are listening
A sentence that reads well on a page can sound crowded when spoken. Keep each sentence focused on one idea, use familiar words, and read the script aloud before generating audio.
For example, “The platform facilitates an accelerated onboarding process” is harder to follow than “The platform helps new users get started faster.” The shorter version is easier to follow because listeners have fewer words to process.
Flag names and acronyms that need special attention. Do the same for dates, prices, and technical terms. Speech software has to turn written forms into spoken words, and a mispronunciation can distract listeners even when the rest of the narration sounds convincing. The W3C speech synthesis standard describes controls for pronunciation, pauses, and delivery.
Direct the pacing and emphasis
Listen for whether the voice stresses the right words and pauses between complete thoughts. Research on current speech systems finds that intonation and pauses between phrases help speech sound natural, even when individual words sound realistic (prosody research).
Start by rewriting an awkward sentence. Then correct pronunciation or adjust the voice and pace. Add pauses only when they help separate ideas, give viewers time to look at a screen, or set up a reveal. A pause after every phrase makes narration choppy; too little space can rush viewers through an important visual.
Listen to the recording while watching the video. A line that sounds fine on its own may play before the screen changes or leave viewers too little time to read an on-screen label.
Make a narrated video with explainroo
explainroo is a free, open-source kit under the MIT license. It works with coding agents that can run shell commands, including Claude Code, Codex, Pi, OpenCode, and Gemini CLI. It works best with Claude Code.
To get started, paste this prompt into your coding agent, replacing the bracketed topic:
Make me a short explainer video about [your topic]. Use explainroo for it: clone the explainroo repository, read its AGENTS.md and follow the steps.
The agent sets up explainroo and creates the video. It checks the video before giving you an MP4 file. It writes script.md, which contains the narration, and scenes.js, which describes what appears on screen. You can ask the agent to revise either file when the result needs work.
For narration, explainroo uses Kokoro, an open voice model with natural American and British English voices. It does not require an account or API key, and its narration is English only. Whisper, a speech-recognition model, marks when each word is spoken so the visuals match the narration. A pace setting in video.json adjusts the speed of the voice, pauses, and animations.
The kit also makes background music and sound effects, then lowers the music while the voice speaks. explainroo gives the agent scene stills and a contact sheet to review. It also checks for overlapping or cut-off text and mispronounced words. The agent cannot watch the finished video, so listen and watch it yourself before publishing.
Follow the explainroo documentation for setup details. You can also inspect example videos before choosing a visual style. explainroo makes animated explainer videos and product demos; it does not edit existing camera footage or create talking avatars or live-action footage. It has no drag-and-drop editor. The kit itself costs nothing per video, while the coding agent is a separate service with its own pricing.
Check the finished narration
Play the video from beginning to end. Listen for mispronounced names, repeated or missing words, abrupt cuts, and changes in speaking style. Check that the voice remains clear over music and that viewers have enough time to read the screen. If you use a voice that imitates a recognizable person, get clear authorization first; the FTC warns that voice cloning can enable impersonation scams.
