Unattended live talking-head
Keep the shop on camera after hours
Unattended Live Avatar turns one staff or host portrait and short welcome, product, FAQ, and close scripts into talking-avatar clips a store can loop overnight. It clones or picks a voice, then delivers labeled MP4 clips of up to 15 seconds each, and later nights reuse the same face and voice. Unattended Live Avatar works in Claude Code, Codex, Cursor, and other agents that run Skills.
Install Unattended Live Avatar and tell me when I need to sign in: https://beatra.ai/skills/unattended-live-avatar/install.md
For agents:Install guideMarkdown version
Send it to WorkBuddy, Codex, Claude Code, or another agent to install. Sign in once the first time you use it.GitHub ↗Docs ↗
One face, a night-shift clip set
Keep one stable face
Use a single shop portrait so welcome, product, FAQ, and close look like the same presenter.
Speak in a matching voice
Clone or pick a voice, then drive a clear talking-head for each loop slot.
Hand over a loop set
Deliver labeled welcome, product, FAQ, and close clips the shop can play overnight.
How it is made
Build the night shift around a face the shop already uses
Use one clear shop or host portrait and keep framing, expression, and background restrained from slot to slot.
Use one stable portrait · Keep the same framing · Hold expression to the slot
Write short scripts the room can repeat
Prepare welcome, product, FAQ, and close copy, then review the narration before each talking-head clip.
Keep each slot easy to say · Hear the narration first · Split a long script into sequential clips
Hand the shop a set it can loop
Direct a stable talking-head, then reuse the same likeness and voice the next night.
Label welcome, product, FAQ, and close · Review identity and mouth timing · Reuse the same face and voice next night
Welcome, product, FAQ, and close loops
Overnight shop loops
Keep a talking-head on the night shift without booking a host.
Product and FAQ clips
Answer the usual questions and show the usual items as separate loop slots.
Open and close the room
Start with a welcome and end with a close the stream can repeat.
Inputs and outputs
Input · Image — One shop or host portrait
- Formats
- JPG, PNG
- Specifications
- 1 image per request
Input · Text — Short welcome, product, FAQ, and close scripts
Input · Audio — Voice sample to clone the shop voice (optional)
- Formats
- MP3, WAV
- Specifications
- 10–300 s; up to 20 MB each
Output · Video — Labeled talking-avatar clips for welcome, product, FAQ, and close, ready to loop
- Formats
- MP4
- Specifications
- up to 15 s
Output · Audio — Narration for each part, in one shop voice
- Formats
- MP3
| Input or output | Type | Formats | Specifications |
|---|---|---|---|
| Input | Image — One shop or host portrait | JPG, PNG | 1 image per request |
| Input | Text — Short welcome, product, FAQ, and close scripts | — | — |
| Input | Audio — Voice sample to clone the shop voice (optional) | MP3, WAV | 10–300 s; up to 20 MB each |
| Output | Video — Labeled talking-avatar clips for welcome, product, FAQ, and close, ready to loop | MP4 | up to 15 s |
| Output | Audio — Narration for each part, in one shop voice | MP3 | — |
Specifications last checked · Sources: Model catalog · Documentation · Terms of Service
Questions
What do I need to start?
One clear staff or host portrait and short scripts for the parts you want to loop. Later nights only need the next scripts.
Can later nights reuse the same face and voice?
Yes. Each new script becomes the next loop clip, made from the same portrait and voice.
What is this for?
Unattended or overnight talking-head loops: welcome, product, FAQ, and close.
Related Skills
