Bytedance Omnihuman API

Omnihuman
V1.5

Endpoint: POST https://fal.run/fal-ai/bytedance/omnihuman Endpoint ID: fal-ai/bytedance/omnihuman

Try it in the Playground

Run this model interactively with your own prompts.

Quick Start

import fal_client

def on_queue_update(update):
    if isinstance(update, fal_client.InProgress):
        for log in update.logs:
           print(log["message"])

result = fal_client.subscribe(
    "fal-ai/bytedance/omnihuman",
    arguments={
        "image_url": "https://storage.googleapis.com/falserverless/example_inputs/omnihuman.png",
        "audio_url": "https://storage.googleapis.com/falserverless/example_inputs/omnihuman_audio.mp3"
    },
    with_logs=True,
    on_queue_update=on_queue_update,
)
print(result)

Input Schema

image_url

string

required

The URL of the image used to generate the video

audio_url

string

required

The URL of the audio file to generate the video. Audio must be under 30s long.

Output Schema

video

File

required

Generated video file

duration

float

required

Duration of audio input/video output as used for billing.

Input Example

{
  "image_url": "https://storage.googleapis.com/falserverless/example_inputs/omnihuman.png",
  "audio_url": "https://storage.googleapis.com/falserverless/example_inputs/omnihuman_audio.mp3"
}

Output Example

{
  "video": {
    "url": "https://storage.googleapis.com/falserverless/example_outputs/omnihuman_output.mp4"
  }
}

Endpoint: POST https://fal.run/fal-ai/bytedance/omnihuman/v1.5 Endpoint ID: fal-ai/bytedance/omnihuman/v1.5

Try it in the Playground

Run this model interactively with your own prompts.

Quick Start

import fal_client

def on_queue_update(update):
    if isinstance(update, fal_client.InProgress):
        for log in update.logs:
           print(log["message"])

result = fal_client.subscribe(
    "fal-ai/bytedance/omnihuman/v1.5",
    arguments={
        "image_url": "https://storage.googleapis.com/falserverless/example_inputs/omnihuman_v15_input_image.png",
        "audio_url": "https://storage.googleapis.com/falserverless/example_inputs/omnihuman_v15_input_audio.mp3"
    },
    with_logs=True,
    on_queue_update=on_queue_update,
)
print(result)

Input Schema

prompt

string

The text prompt used to guide the video generation.

image_url

string

required

The URL of the image used to generate the video

audio_url

string

required

The URL of the audio file to generate the video. Audio must be under 30s long for 1080p generation and under 60s long for 720p generation.

turbo_mode

boolean

default:"false"

Generate a video at a faster rate with a slight quality trade-off.

resolution

ResolutionEnum

default:"1080p"

The resolution of the generated video. Defaults to 1080p. 720p generation is faster and higher in quality. 1080p generation is limited to 30s audio and 720p generation is limited to 60s audio. Default value: "1080p"Possible values: 720p, 1080p

Output Schema

video

File

required

Generated video file

duration

float

required

Duration of audio input/video output as used for billing.

Input Example

{
  "image_url": "https://storage.googleapis.com/falserverless/example_inputs/omnihuman_v15_input_image.png",
  "audio_url": "https://storage.googleapis.com/falserverless/example_inputs/omnihuman_v15_input_audio.mp3",
  "turbo_mode": false,
  "resolution": "1080p"
}

Output Example

{
  "video": {
    "url": "https://storage.googleapis.com/falserverless/example_outputs/omnihuman_v15_output.mp4"
  }
}

ByteDance’s OmniHuman model generates audio-synchronized videos from a single reference image at $0.14 per second of output. With specialized lip-sync precision, the model trained on 18,700 hours of human motion data to maintain tight correlation between audio input and character movement. Built for content creators who need realistic talking head videos without motion capture equipment. Use Cases: Social Media Content | Product Demos with Presenters | Educational Video Production

Performance

At $0.14 per second, OmniHuman sits in the mid-range for image-to-video generation on fal, trading cost for specialized audio-sync capabilities that generic models don’t prioritize.

Metric	Result	Context
Audio Sync Quality	Tight emotion/movement correlation	Trained on 18,700 hours of human motion data
Max Audio Duration	30 seconds	Hard limit enforced at API level
Cost per Second	$0.14	Billed on actual audio/video duration
Output Quality	High-fidelity video	Specialized for human figure animation
Related Endpoints	OmniHuman v1.5, Seedance Pro, Seedance Lite	ByteDance family variants for different quality/cost tradeoffs

Audio-First Video Generation

OmniHuman flips the standard image-to-video workflow by making audio the primary control signal rather than text prompts or motion parameters. Where most models animate based on text descriptions or keyframes, this architecture analyzes audio waveforms to drive facial expressions, lip movements, and body language simultaneously. What this means for you:

Natural speech synchronization: Upload any audio file under 30 seconds and get matching lip movements without manual keyframe adjustment or phoneme mapping
Emotion-driven animation: The model interprets audio tone and pacing to generate corresponding facial expressions and body gestures, not just mouth shapes
Single-image input: Start with one reference photo rather than building character rigs or providing multiple angles
Production-ready output: High-fidelity video maintains visual consistency across the full duration without drift or quality degradation

Technical Specifications

Spec	Details
Architecture	OmniHuman
Input Formats	Single image (JPEG, PNG, WebP, GIF, AVIF) + audio file (MP3, OGG, WAV, M4A, AAC)
Output Formats	MP4 video
Max Audio Duration	30 seconds
License	Commercial use allowed via fal partnership

API Documentation | Quickstart Guide | Enterprise Pricing

How It Stacks Up

Bytedance OmniHuman v1.5 (

0.14/sec) – [OmniHuman v1.5](https://fal.ai/models/fal-ai/bytedance/omnihuman/v1.5) offers enhanced audio processing and **improved motion quality** at the same

0.14 per second price point. The original OmniHuman remains viable for projects where the current quality threshold meets requirements without needing v1.5’s refinements. Seedance 1.0 Pro ($0.14/sec) – OmniHuman prioritizes audio-driven human animation with specialized lip-sync capabilities. Seedance Pro trades audio control for broader creative motion generation across any subject type, ideal for product animations or scene transitions where audio sync isn’t critical. Seedance 1.0 Lite (

0.08/sec) – [Seedance Lite](https://fal.ai/models/fal-ai/bytedance/seedance/v1/lite/image-to-video) delivers 43% cost savings (

0.08 vs $0.14 per second) by simplifying motion generation without audio processing. Best for budget-conscious projects that can handle reduced quality or don’t require speech synchronization.

Limitations

resolution restricted to: 720p, 1080p

Video Generation

Image Generation

Audio

Vision

3D

Bytedance Omnihuman API

Try it in the Playground

Quick Start

Input Schema

Output Schema

Input Example

Output Example

Try it in the Playground

Quick Start

Input Schema

Output Schema

Input Example

Output Example

Performance

Audio-First Video Generation

Technical Specifications

How It Stacks Up

Limitations

Video Generation

Image Generation

Audio

Vision

3D

Try it in the Playground

​Quick Start

​Input Schema

​Output Schema

​Input Example

​Output Example

Try it in the Playground

​Quick Start

​Input Schema

​Output Schema

​Input Example

​Output Example

​Performance

​Audio-First Video Generation

​Technical Specifications

​How It Stacks Up

​Limitations

Quick Start

Input Schema

Output Schema

Input Example

Output Example

Quick Start

Input Schema

Output Schema

Input Example

Output Example

Performance

Audio-First Video Generation

Technical Specifications

How It Stacks Up

Limitations