Bring a single portrait to life
Combine an avatar, a script, and a voice. Set up the avatar once and change the script for each video. Choose an AI avatar to stay off camera, or clone your own voice.
See the results first
Every video below was generated with xiats. All avatars are AI-generated, with no live filming.
Three steps to a video
Combine an avatar, a script, and a voice. Set up your avatar once; change only the script for each new video.
Choose an avatar
Upload a video of yourself facing the camera, or generate an AI character if you prefer to stay off camera.
Add a script
Extract speech from platform share links, upload a video for transcription, or ask AI to write a script.
Choose a voice and generate
Pick a library voice or clone your own. The system adds voiceover and lip sync to produce the finished video.
- Only your own assetsUpload an authorized avatar or generate one with AI. The Platform does not collect other people's portraits for you.
- Temporary output storageGenerated files are temporary and retained for about 24 hours. Download them promptly.
- Credentials managed on the serverPlatform credentials are maintained in the backend and persisted in the database, without hardcoding or exposure.
- Two key modesPlatform mode uses official keys out of the box. BYOK uses your key and bills your cloud account.
Why use xiats for talking avatars?
Create more than lip sync: manage your avatar, voice, script, and cost in one place.
No need to appear on camera
Four presets are available: an articulate female presenter, a professional male narrator, an energetic creator, and a 3D virtual character. All are AI-generated without real-person portrait licensing.
Clone your voice
Use a library voice or clone your own. Pair the same avatar with different voices and choose your presentation style.
No need to write from scratch
Extract speech from Xiaohongshu, Douyin, Bilibili, or Kuaishou links, or transcribe an uploaded video. AI can rewrite it into a more natural spoken script.
Know the cost before generation
Voiceover is billed by character count and lip sync by output duration. A confirmation dialog shows “Voiceover ¥X + Lip sync ¥Y = Approximately ¥Z” before generation, with no hidden charges.
Full automatic refunds on failure
Failed jobs are not charged. Reserved funds return to your wallet automatically, without contacting support.
One avatar, many videos
Record your avatar footage once. Change the script for each new video and generate in minutes, without setting up lighting and cameras again.
What people create
When the same person needs to deliver different messages repeatedly, talking avatars save time.
Product presentations
Create a presentation for each product using the same avatar, and prepare a batch for launch day.
Daily educational content
Turn topics into scripts and let an avatar present them, so daily publishing does not depend on filming.
Company communications & training
Keep product introductions, onboarding, and internal announcements consistent. Update the script to update the video.
Test different versions
Try variations in script and voice to see which performs best, paying only for the generated versions.
How avatar generation is priced
- Voiceover: Billed by the script's character count.
- Lip sync: Billed by actual output duration, approximately equal to the voiceover length, with a minimum charge.
- Before generation, a dialog shows “Voiceover ¥X + Lip sync ¥Y = Approximately ¥Z” and your current balance. You are charged only after confirmation.
- Failed generation is fully refunded to your wallet automatically, without contacting support.
See the workspace price list for current rates. Sign in and open Talking avatars for a live quote. Generation is charged by usage from your wallet, independently of membership. View subscriptions。
Frequently asked questions
Create your first talking video
Upload avatar footage or choose an AI avatar, add a script, and select a voice.