The short answer: whiteboard videos can be narrated two ways: record your voice segment by segment with a microphone, or connect your own TTS API in the AI Voice tab and generate the narration per segment (with reference-audio voice cloning); then translate the script and canvas text into 13 languages, switch languages, regenerate the voiceover and record again; one canvas, many language versions.
The two narration modes
During segment recording, each segment's narration comes from one of two places:
| Mode | Workflow | Character |
|---|---|---|
| Mic recording | Record with your microphone segment by segment, reading from the teleprompter | Most personal; tweak as you go |
| AI voiceover | Generate TTS from the script in the AI Voice tab, then record while the timeline plays | No need to speak, multi-language, easy re-records after script changes |
The two mix freely per segment: record the main body with your voice, use AI for the intro and outro.
The recording flow itself (segments, slide linking, burned subtitles, export) lives in the recording guide; this article focuses on voice and languages.
Setting up your own TTS with voice cloning
The AI Voice tab starts with a one-time setup: your own TTS service:
- Endpoint and auth: works with common TTS API formats; request field names are customizable;
- Model and language code: filled per your service;
- Reference audio + reference text: provide a sample clip and its transcript for voice cloning, so the TTS reads your script in that voice.
In other words, there is no built-in voice library; your TTS service defines the voices. Generated clips cache in your browser, so re-recording the same script doesn't regenerate audio.
One card per segment: generate TTS, preview, regenerate, or generate all segments concurrently
Per-segment generation and narration-mode recording
The voiceover is organized per segment, mapped to slides:
- Generate per segment: each segment card generates TTS, plays it back, and regenerates on demand;
- Generate all: concurrent generation for every segment, cached locally;
- Narration-mode recording: while recording, the TTS timeline plays automatically and the canvas switches per segment, producing a complete narrated video in one take;
- Export: same as normal recording, straight to MP4.
Once the script is parsed into segments, voiceover, subtitles and slides share one structure
Subtitles and voiceover come from the same script: subtitles are burned from the estimated timing, the voiceover is the TTS reading, so content matches by construction; timing alignment is tuned via the subtitle speed setting.
Translating into 13 languages
Both the narration script and the canvas text elements can be AI-translated:
- Scope: both / script only / canvas text only;
- Target languages: 13 (Chinese, English, German, French, Spanish, Japanese, Korean and more);
- Translation service: also self-configured; fill in your LLM API (OpenAI Chat Completions and Anthropic Messages protocols both supported);
- Style presets: conversational / formal / casual;
- Translations are editable per segment and stored alongside the original language.
Translation applies to the narration script and canvas text; shapes and structure stay untouched
The switch-and-re-record workflow
One canvas, many language versions; the key is re-voicing after translation:
- Translate: pick a target language and translate the script and canvas text (editable per segment);
- Switch language: the project switches, and canvas text and script follow;
- Regenerate the voiceover: TTS regenerates per segment from the new-language text;
- Re-record: run narration-mode recording again and export the target-language MP4.
The original language's data stays in local storage; switch back and re-record anytime.
vs a manual voiceover workflow
| Manual voiceover / post tools | Whiteboard studio | |
|---|---|---|
| TTS | Find a separate TTS tool, then align audio by hand | One click per segment, auto-aligned to segments |
| Multi-language | Rewrite, re-record and re-align per language | One-click script translation; switch and regenerate |
| Subtitles | A subtitle app for timing | Burned automatically from the same script |
| Voice | Fixed voices or another purchase | Your TTS service, with voice cloning |
| Data | Assets scattered across tools | Everything lives in one local project |
In one line: manual workflows win on swapping tools at every step, and the studio wins on "change the script once; voiceover, subtitles and translation all follow".
Frequently asked questions
Does AI voiceover need my own TTS service?
Yes. There is no built-in voice library; fill in your own TTS API in settings (endpoint, auth, model and language code), with customizable request field names and common API formats supported.
Does it support voice cloning?
Yes. Provide a reference audio clip and its transcript, and the TTS reads your script in that voice; configured in settings.
Which languages are supported?
Translation targets 13 languages (Chinese, English, German, French, Spanish, Japanese, Korean and more); TTS languages depend on your service, set via the language code.
Can mic recording mix with AI voiceover?
Yes, per segment: some segments mic-recorded, some TTS-generated, merged into one export.
Do subtitles stay in sync with the voiceover?
Both come from the same script, so content matches by construction; subtitle timing is estimated from speed; adjust the subtitle speed in settings to better match the pacing.
Does translation change the canvas graphics?
No. Translation only touches the narration script and the text content of canvas elements; shapes, positions and structure stay unchanged.
What does it cost?
AI voiceover and translation call your own TTS and LLM services, so the cost lands on your service side; canvas recording and watermark removal involve Eyesme's paid upgrade.
Where is the data stored?
Script, translations, segment videos and voiceover audio live in your browser's local storage and are never uploaded.
Features and details reflect the studio as of November 12, 2026 and may evolve; product screenshots are from the current interface.
Ready to make one canvas speak many languages? Open Draw Studio, record the original with the recording guide, then use AI voiceover and translation for every language version.




