OmniHuman-1
Audio-driven talking-head generation: one portrait plus audio becomes a lifelike speaking video.
About
OmniHuman-1, published by ByteDance's Intelligent Creation team in early 2025, turns a single portrait image and an audio clip into a talking-head video with synchronized lips, expressions and head motion. It needs no text prompt — the audio drives the whole performance, handling speech, singing and stylized or illustrated faces. A follow-up, OmniHuman-1.5, improved lip-sync accuracy and robustness on non-photorealistic portraits.
ByteDance never open-released weights or an official API, so access runs through third-party hosts (fal, RunComfy, Astorie and similar) that serve the model behind their own APIs. Despite that, it has become the default single-shot route for portrait animation in creator tooling: feed it a headshot and a voiceover and it returns a video whose length tracks the audio. Anyone using it should be aware the usual deepfake-misuse concerns apply.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
No published benchmarks found
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 6/10
Scored 6/10 (TIER metrics, metrics-v1). Evidence: ByteDance research release (early 2025) with v1.5 follow-up; broad adoption across third-party video APIs (fal, RunComfy, Astorie); consistently described as the strongest single-shot portrait-animation route in creator tooling. Strengths: state-of-the-art lip-sync quality, wide real-world usage, simple image+audio interface. Docked for: (1) no open weights and no official ByteDance API — all access is via third parties; (2) no published benchmark figures. License: proprietary (no public license; ByteDance has not released weights).
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 7.0/10
- Maintenance · recency, releases, issue load
- 6.0/10
- Openness · license + docs
- 2.0/10
- Manual review · sanity read + red-flag check
- 6.0/10
- · ByteDance Intelligent Creation, early 2025; v1.5 follow-up with better lip-sync
- · Widely served via third-party APIs (fal, RunComfy, Astorie)
- · No open weights, no official ByteDance API, no published benchmarks