Lore
Directing the avatar with an action prompt
# Directing the avatar with an action prompt
In Spiel's UGC builder, the script only controls what the avatar says. A separate plain-language action prompt controls what the avatar does while talking. The source's line is: "You can actually tell your avatar what you want it to do in the video."
What it can direct
- Standing up
- Taking a hat off
- Interacting with the product, for example holding it or presenting it
- Drinking from a bottle
How to apply it
- Finish the script step first. See Custom GPT for viral-optimized scripts and Spiel UGC builder: Easy mode vs Custom mode.
- Type the action as an ordinary sentence in the instruction field. The demo prompt was: 'she speaks directly to the camera whilst presenting the bottle. At the end of the video, she takes a drink from the bottle'.
- Click generate and wait a couple of minutes.
Why it works
- The action prompt gives the product a moment on screen. The product reference is set up in Product-in-hand image upload.
- A timed beat such as 'at the end' gives the clip a payoff. That makes it feel like a real creator demonstrating the product, not a talking head reading a script.
- It sits at the video stage of Turning image variations into talking UGC video ads and keeps the character and product from the source image.
The action prompt does nothing for the audio. The voice still needs the fixes in Audio realism as the last step that makes AI UGC pass.