I script and produce all our release note videos in Synthesia. This template saves me hours. It's a direct prompt for the AI avatars.
**Base Prompt Structure:**
```
Avatar: [Choose a consistent one, e.g., Anna]
Background: [Your branded template]
Script:
- (Upbeat) "Hello everyone, and welcome to version [X.Y.Z]."
- (Calm, direct) "This release focuses on three key areas: [Area 1], [Area 2], and [Area 3]."
- (Slight head nod) "First, [Feature 1]. This allows you to [clear user benefit]. Here's a quick look."
- *Insert screen recording clip or visual*
- (Matter-of-fact) "Next, we've improved [Feature 2] to fix [specific pain point]."
- *Insert screen recording clip or visual*
- (Slight smile) "Finally, behind the scenes, we've enhanced performance. [Metric] is now [X]% faster."
- (Warm, concluding) "That's all for this update. Update now and let us know what you think. Thanks for watching."
```
**Key Findings from Testing:**
* Keep sentences short. Avatars handle pauses better than long, complex sentences.
* Use the (direction) tags. They significantly impact delivery more than word choice alone.
* Always pair verbal claims with a visual (screen recording, text callout, UI image). The avatar's "talking head" alone tests poorly for retention.
* For version numbers and metrics, use Text/Number assets, not spoken-only. Viewers remember them better.
This template yields a consistent, scannable 60-90 second video. It's efficient.
Benchmarks don't lie.
This is a really clean template. I like how you've structured the script around the (direction) tags, those small cues make a big difference in how the avatar lands a line. One thing I've noticed in testing similar prompts is that the visual pairing works best when the screen recording clip is explicitly cued in the script timing, like a countdown or a frame marker, so the avatar doesn't wrap up the sentence before the clip appears. Have you found any specific way to sync the "insert visual" instruction with the synthesia timeline to avoid awkward pauses?
Reviews build trust.
That timing question is so key. I've been experimenting with the same thing, and honestly, I gave up on relying purely on script cues. I now treat the Synthesia prompt more as a storyboard and handle the final sync in the video editor.
I drop the avatar video and the screen clips into separate tracks, then nudge the clips to start right after the avatar says "Here's a quick look." It's an extra step but way more reliable. Does the Synthesia timeline have any native marker feature for that yet?