AI avatar and lip-sync spec builder
AI avatar tools — HeyGen, Synthesia, D-ID — handle lip-sync automatically, but they don’t plan your video for you. A flat result usually comes from no thought given to framing, expression timing, and runtime. This builder turns a script into a clean shot specification: it estimates spoken runtime from your word count and pacing, and assembles your avatar style, camera angle, background, and expression notes into a spec you can follow when configuring any avatar tool.
How it works
Runtime estimation. The builder divides your script’s word count by a words-per-minute pace you choose — slow (around 110 wpm), natural (around 135 wpm), or brisk (around 160 wpm). Avatar platforms like HeyGen and Synthesia bill by video duration, so knowing the runtime before you start avoids cost surprises mid-project.
Framing decisions. Every avatar platform exposes the same core framing options. The builder collects them in one place:
- Avatar style: professional, casual, or stylized.
- Camera angle: eye-level (most natural), slight high or low angle, or close-up for emphasis.
- Background: solid brand colour, office environment, custom image upload, or blurred location.
Expression timing. Avatar platforms sync lips automatically — but not emotion. The builder lets you annotate where the delivery should shift: a warm smile on the opening, a serious tone for the key point, a deliberate pause before the call to action. Without these notes a 90-second talking-head risks looking robotic the whole way through.
Platform compatibility
The spec maps directly to settings in the most-used avatar tools:
| Setting | HeyGen | Synthesia | D-ID |
|---|---|---|---|
| Avatar style | Avatar selection | Presenter style | Actor selection |
| Camera angle | Framing preset | Shot type | — |
| Background | Scene background | Background | Background |
| Expression/emotion | Expression control | Tone | — |
The tool generates a plain-text specification you paste or follow in whichever platform you use — it does not connect to any API or generate the video itself.
Tips for natural avatar videos
- Write for the ear. Short sentences and contractions lip-sync more naturally than dense written prose. Reading time and speaking time differ — always base duration on the word count tool, not a read-through.
- Keep clips under roughly 90 seconds. Attention drops fast on talking-heads; split a 3-minute script into 2–3 scenes with different framings.
- Vary expression every 2–3 sentences. A single fixed expression throughout is the clearest giveaway that a video is AI-generated.
- Match background to context. A blurred office reads as professional. A solid brand colour reads as an advertisement. A custom location image reads as on-site. Choose deliberately rather than leaving it at the default.
- Proof-listen before rendering. Paste the script into a text-to-speech tool first — mispronounced names or acronyms will need phonetic spelling in the avatar platform’s script field.