Video generation
Coming soon

OmniHuman 2.0

by ByteDance

ByteDance's audio‑driven human video model.
One photo, one track, five minutes in one take.

OmniHuman 2.0

Key features

Make this photo speak my voiceover in one continuous takeStart creating

Specs are BytePlus's own, from its October 15, 2026 launch post. OmniHuman 2.0 is not on Morphic yet, so the clips on this page were made on Morphic with Veed Fabric, which turns one photo and one audio track into a talking video up to 5 minutes long today.

Technical specifications

5 min

One continuous shot, per BytePlus

60 s

The single-shot cap before 2.0

Photo + audio

One image of the performer and one track

Any track

Speech, music or dialogue

Use cases

A gardening instructor in a linen shirt teaching beside raised vegetable beds

Courses and lessons

A full lesson in one take, so the instructor looks the same in minute four as in minute one and nothing needs cutting together.

A chef in a white jacket talking to camera in a kitchen with copper pans

Explainers and how-to videos

Record or generate the voiceover first, then let one photo deliver it, with the pauses and emphasis that keep a viewer watching.

A singer in a burgundy dress holding a lit match under warm stage light

Music performance clips

Give it a sung track and the face carries the long notes, the breaths and the feeling of the song, not just the words.

A clay fox character in a blue scarf sitting at a craft table with a paper boat

Mascots and characters

Give a brand mascot, a clay character or an illustrated hero a voice for ads, kids' content and channel intros.

A young woman with a blonde bob smiling into the camera on a cobblestone street

Creator and social videos

Vertical talking clips for reels and shorts from a single selfie-style photo and a voice track, ready to post.

An elderly man in a green cardigan by a window overlooking a winter river

Stories and podcasts

Long spoken stories keep their rhythm, from the quiet build to the punchline, without the performer resetting every minute.

Overview

OmniHuman 2.0 is ByteDance's audio-driven human video model, released through BytePlus on October 15, 2026. It turns one photo and one audio track into a video of that person performing the track, and it reads the tone, the emotion and the words in the audio rather than only matching mouth shapes. A single shot can now run up to 5 minutes, up from 60 seconds, with no stitching (source).

What OmniHuman 2.0 does differently

Most talking-photo models move the mouth to the sound. OmniHuman 2.0 listens for how a line is said, so a whispered bedtime story, a shouted cue from a coach and a held note in a song each come out as different performances from the same kind of input. The longer single take matters as much: a lesson, a story or a full song runs as one shot, so there is no join where the face, the light or the background can change.

OmniHuman 2.0 on Morphic

OmniHuman 2.0 is coming to Morphic soon. Until then, Veed Fabric does the same job: open Copilot, attach a photo and an audio track, and ask for a talking video. Every clip on this page was made that way. For how to get the best result, see the OmniHuman 2.0 guide.

Simple pricing

Get started for free today, with the option to upgrade or cancel anytime.

Basic

$19/ month
billed as $0 per year

2400 monthly credits

1 user only

All models

Workflows

Standard

$24/ month
billed as $0 per year

3625 monthly credits

1 user only

All models

Workflows

Pro

$45/ month
billed as $0 per year

6350 shared monthly credits

1 user

+ up to 4 more at extra cost

All models

Workflows

Max

$170/ month
billed as $0 per year

24650 shared monthly credits

1 user

+ up to 9 more at extra cost

All models

Workflows

Enterprise

For higher limits

Custom

pricing and billing terms

High-volume credits
Custom seat limits
All models
Workflows
Pricing Gradient

Free

For playing around

$0

forever free

Up to 20 credits
1 user only
Limited models
Workflows
Compare plan details

FAQs

What is OmniHuman 2.0?
OmniHuman 2.0 is ByteDance's audio-driven human video model, released through BytePlus on October 15, 2026. You give it one photo and one audio track, speech, music or dialogue, and it generates a video of that person performing the track, with the tone, emotion and words in sync, in one continuous shot up to 5 minutes long.
When is OmniHuman 2.0 coming to Morphic?
OmniHuman 2.0 is coming to Morphic soon. Until it arrives, Veed Fabric does the same job on Morphic today: it turns one photo and one audio track into a talking video up to 5 minutes long. Every clip on this page was made with it.
What is new in OmniHuman 2.0?
Three things, according to BytePlus. It reads the tone, emotion and words in the audio and performs them, beyond matching mouth shapes. A single shot can run up to 5 minutes, up from 60 seconds, with no stitching. And it costs less to run than the version before it.
How long can an OmniHuman 2.0 video be?
Up to 5 minutes in one continuous shot, according to BytePlus. The previous version stopped at 60 seconds, so longer videos had to be generated in pieces and joined, which could make the face or background jump at each join.
Can OmniHuman 2.0 make a photo sing?
Yes. BytePlus says it takes any track, including music, and performs it. Give it a sung track with the voice clear and up front, and the face follows the lyrics, the held notes and the breaths.
How is OmniHuman 2.0 different from lip sync?
A lip sync model changes the mouth in a video that already exists. OmniHuman 2.0 starts from a single photo and generates the whole performance, face, head and body, from the audio. Use lip sync to re-voice footage you have, and an audio-driven model when all you have is a picture.
Why does the background move in some OmniHuman videos?
Some early viewers of BytePlus's launch demo noticed the background shifting. Audio-driven models tend to animate more of a busy photo than you intended, including other faces. A plain, uncluttered background and one clear face in the photo give the model less to move. The OmniHuman 2.0 guide has a side-by-side test.
Were the clips on this page made with OmniHuman 2.0?
No. OmniHuman 2.0 is not on Morphic yet, so every clip was made on Morphic with other models: the photos with Krea 2 Large, the voices with ElevenLabs v4, the song with ElevenLabs music, and the talking video with Veed Fabric.

More about OmniHuman 2.0

Guides, prompts, and comparisons for OmniHuman 2.0.

OmniHuman 2.0 guide

How to use OmniHuman 2.0: pick the right photo, prepare the audio, run a five-minute single take and stop the background from flickering.

Other models

Explore the rest of the Morphic model catalog.

See all models →

Mirelo SFX 2.0

Mirelo AI

Mirelo's video-to-sound model. New sound for your video, same dialogue.

Fish Audio Drama 3

Fish Audio

Fish Audio's most expressive voice model. Direct the performance, down to the word.

Vidu Q4

Shengshu Technology

Shengshu's latest video model. Up to 12 reference images and 3 voices held consistent across one clip, with native audio from 540p to 4K.

Nano Banana 2.1

Google DeepMind

Google DeepMind's new Flash image model. Better design, cleaner edits, steadier characters.