Duix-Avatar API — Audio+Video Driven Digital Human
Duix-Avatar is KeyoAPI’s audio+video driven talking-avatar API: feed a face video plus driving audio, get lip-sync and expressions back as an async render job.
Overview
Duix-Avatar is KeyoAPI’s audio+video driven talking-avatar API: feed a face video plus driving audio, get lip-sync and expressions back as an async render job. Built for course creators, support explainers, and product demos that need a speaking face without a custom Unreal pipeline. Digital humans used to mean custom game engines. HTTP async APIs let SaaS teams productize avatars faster. KeyoAPI collocates TTS, STT, and avatar models so you can build the full loop on one relay key. Per-second pricing (~$0.0068) maps cleanly to video length sold to end users. Cap max duration in your product to control worst-case invoices. Training videos, localized spokesperson clips, and in-app guides. Pair with InfiniteTalk when you start from a still image plus audio. Async endpoint POST /v1/async/videos/audio-video-to-video with model=Duix-Avatar. Keys: /pricing/Duix-Avatar. Submit the job, then poll /v1/task/{id} for the rendered clip. Validate face video and voice sample length before submit; show progress UI and do not block HTTP workers. Moderate inputs for deepfake abuse. Cache finished MP4s by (face hash, audio hash). Buyers care about lip-sync quality and queue time—log both. Validate face frontal angle and audio SNR before queueing; show ETA from queue depth. Abuse prevention matters as much as render quality—rate-limit and log. Generate speech with IndexTTS-2, CosyVoice3, or Qwen3-TTS, then drive Duix-Avatar on the same key.
Quick start
Base URL: https://www.keyoapi.xyz/v1
Endpoint: POST /v1/async/videos/audio-video-to-video
Model: model=Duix-Avatar
curl https://www.keyoapi.xyz/v1/async/videos/audio-video-to-video \
-H "Authorization: Bearer YOUR_API_KEY" \
-F model=Duix-Avatar \
-F audio=@./voice.wav \
-F video=@./face.mp4
# then: GET /v1/task/{id}
Auth uses the same Bearer API key as chat. Full notes: Keyo docs.
FAQ
Is generation realtime?
Duix-Avatar runs as an async job—submit audio/video drivers, then fetch the rendered avatar clip.
How do I supply voice?
Generate speech with IndexTTS-2, CosyVoice3, or Qwen3-TTS (or upload audio), then drive Duix-Avatar for lip-sync and expressions.
Billing unit?
About $0.0068 per second of output—see /pricing/Duix-Avatar.