When an AI-made video puts text on screen, two problems are easy to miss: text can overlap or get cut off, and captions or visuals can appear at the wrong time. For narrated explainer videos, explainroo is the best fit for checking these problems. It checks layouts for overlapping or cut-off text and checks speech for mispronounced words.
A subtitle is timed to spoken words. On-screen text, such as a heading or label, is part of a scene. Check both against the audio and video: even readable text fails if it appears too early, too late, or covers important content. Subtitle guidance recommends keeping each line onscreen long enough to read. At least one second per line is a practical baseline. Place subtitles where they do not block important visuals (subtitle timing guidance).
Check both the layout and the timing
Watch the whole video, then inspect frames where text changes or the narration introduces a new point. At normal speed, check whether each visual appears with the relevant narration and whether scenes change at a sensible moment. Pause when text looks crowded, overlaps, or extends beyond the frame.
A contact sheet, or page of small frame previews, helps reveal problems that repeat across a video. Use still frames to check whether text is clipped or crowded, and playback to check whether visuals appear with the narration. General guidance on reviewing AI-made video recommends checking visual and audio signals alongside motion and file signals, since one glitch alone does not establish a broader problem (AI video review signals).
Listen while reading the captions. Check whether a caption appears before the speaker says the words, stays onscreen after the phrase ends, or changes too quickly to read. If a whole subtitle line is early or late by the same amount, shift its timing. If the mismatch grows through the video, stretch or shorten the timing instead. Automated synchronization can help, but a person still needs to review speech recognition and sync results, especially when the recording is noisy or the speaker has a strong accent (subtitle synchronization guidance).
Check an explainroo video
explainroo is a free, open-source kit for coding agents to create narrated explainer videos and product demos. The agent writes script.md for the narration and scenes.js to describe what each scene draws. explainroo uses Whisper, a speech-recognition tool, to mark when words are spoken so scene elements appear on cue. It also creates stills and a contact sheet, checks layouts for overlapping or clipped text, and checks speech for mispronounced words.
To create and check a video, give a coding agent that can run shell commands this prompt, 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, makes and checks the video, then gives you an MP4 file. Start with the layout check. Then use the scene stills and contact sheet to find crowded or clipped text. Play the MP4 while following the narration to catch visuals that appear at the wrong word or captions that do not match the speech. explainroo supports word-by-word captions in tall, 4:5, and square formats. Its wide format is intended for YouTube.
When you find a problem, give the agent a specific change request, such as moving a label away from another block of text or showing a diagram when a particular spoken word begins. The agent can revise the scene description or narration and run the checks again. For details on supported formats and setup, see the explainroo documentation.
Fix the cause before exporting again
If two labels overlap, revise their placement or simplify the text so each has room. If text is cut off, adjust its position or layout and rerun the layout check. If a visual appears at the wrong point in the narration, ask the agent to align it with the intended spoken word. If the narration itself contains a mispronounced word, correct the script and check the new audio.
Then watch the exported MP4 from beginning to end. Automated checks help find problems, but playback confirms that the text is readable and the visuals match the narration. For a separate subtitle file, a timing tool such as AutoSubSync can synchronize subtitles automatically or apply a manual time shift. Watch the result with the audio to confirm the timing.
