OmniHuman 2.0
NEWOmniHuman 2.0 — ByteDance's talking-video model. Give it one photo of a person and a voice track of up to 5 minutes, and it makes a video of them speaking it, with lip sync, facial expressions and body movement that follow the audio. Add an optional prompt to direct the scene and camera.
Capabilities
Specifications
Estimated Generation Time
~2-5 min
API Usage
curl https://www.flixly.ai/api/v1/generate \
-H "Authorization: Bearer flx_live_..." \
-H "Content-Type: application/json" \
-d '{
"model": "omnihuman-2",
"prompt": "A beautiful sunset over the ocean",
"type": "REFERENCE_TO_VIDEO"
}'Get your API key from the Developer Portal.
Related Models
View allVeo 3.1
Premium video generation model with native audio
Seedance 2.5
ByteDance's newest video model. Generates up to 30 seconds in a single shot with synchronized audio — dialogue, music, and effects in the same pass — from a prompt, a start frame, or up to 50 image, video, and audio references.
Wan 3.0 Prime
The premium tier of Alibaba's Wan 3.0 — faster turnaround, cinematic motion, and stronger visual continuity. Text, image, and reference modes, native 30-second clips in a single pass, with audio.
MiniMax H3 Max
Post-trained MiniMax H3, ranked #1 for overall quality, prompt understanding, and aesthetics against leading video models. 5-15 second clips in native 480p, 768p or 1080p from a prompt, a start frame, or up to 9 reference images, 3 motion clips and 3 audio clips. Reference-to-video is generally available: up to 2x faster, with much stronger subject preservation.