BytePlus OmniHuman 2.0 is an audio-driven AI video generation model that transforms images and audio into lifelike talking avatar videos. It supports natural lip sync, expressive facial movements, body gestures, and video generation up to 5 minutes long.

Pricing: 14 credits per second ($0.07/s), billed on the output video duration. 720P output, length follows the input audio (4–300 s). High-tier top-ups include +10% bonus credits. Effective pricing is approximately 10% lower than the above rates.
Input

Text prompt (Character Tendency). used for precise control of the scene,movements.camera movements,etc. Supported Languages: Only Chinese, English, Japanese, Korean, Spanish (Mexican), Indonesian.

0/300

Click to upload or drag and drop

Supported formats: MPEG, WAV, X-WAV Maximum file size: 15MB

Audio url. The duration [4,300] s

Click to upload or drag and drop

Supported formats: JPEG, PNG, WEBP, BMP, TIFF, GIF Maximum file size: 30MB; Maximum files: 1

Image URL. Only one image is supported. Aspect Ratio (Width/Height): [0.4,2.5]. Dimension (px): [300,6000]

Whether to enable the watermark. Default: true.

Output
output typevideo

README

Complete guide to using omnihuman-2-0

OmniHuman 2.0 API for AI Talking Avatar Videos

Original image
FeatureOmniHuman 1.5 APIOmniHuman 2.0 API
InputImage + Audio + Optional PromptImage + Audio + Optional Prompt
Audio-Driven AnimationSupportedSupported
Lip SyncSupportedSupported
Facial ExpressionsSupportedSupported
Body GesturesSupportedSupported
Multi-Person AnimationSupportedSupported
Text Prompt ControlSupportedSupported
Maximum Video DurationUnder 60 SecondsUp to 300 Seconds*
Output Resolution720p / 1080p720p*
Best ForShort, High-Resolution Avatar VideosLonger Talking Avatar Videos
Price27 credits/s ($0.135)14 credits/s ($0.07)
OmniHuman 2.0 specifications may vary by API provider and endpoint.

OmniHuman 2.0 API is an audio-driven AI video generation API developed from ByteDance's OmniHuman 2.0 model. It turns character images and audio into talking avatar videos with synchronized lip movements, facial expressions, and body gestures. Developers can integrate OmniHuman 2.0 API through Kie AI to build AI avatar and video creation applications.

Create a key on the Kie AI API Keys page (kie.ai/api-key) and send it as the header Authorization: Bearer YOUR_API_KEY. Then call POST https://api.kie.ai/api/v1/jobs/createTask with model set to omnihuman-2-0 and an input that contains image_urls (a one-element array) and audio_url, plus an optional prompt. The request returns a taskId: pass a callBackUrl to be notified when the video succeeds or fails, or poll GET /api/v1/jobs/recordInfo?taskId=…; on success, resultJson.resultUrls holds the MP4 link. Every parameter is in the OmniHuman 2.0 API docs.

ByteDance OmniHuman 2.0 uses a reference image and an audio file to generate animated character videos. The model analyzes speech, rhythm, and audio cues to create matching lip movements, facial expressions, and body gestures. Developers can also use optional text prompts to guide character actions and performances.

OmniHuman 2.0 API takes one image per request in JPEG, PNG, WebP, BMP, TIFF, GIF, HEIC or HEIF format, under 30 MB, 300 to 6,000 pixels per side and with a width-to-height ratio between 0.4 and 2.5. Audio must be MP3 or WAV, under 15 MB and 4 to 300 seconds long. For complete input requirements, check the Kie AI API documentation.

OmniHuman 2.0 API supports videos from 4 to 300 seconds (5 minutes), with the generated video matching the input audio duration. The model outputs videos at 720p resolution, making it suitable for extended talking avatar videos, presentations, and singing performances. For endpoint-specific limits, refer to the Kie AI API documentation.

OmniHuman 2.0 costs 14 credits ($0.07) per second of output video at 720p, and the video length follows your audio. A 30-second clip costs 420 credits ($2.10) and a full 5-minute (300-second) video costs 4,200 credits ($21). OmniHuman 1.5 is billed at 27 credits ($0.135) per second. Prices may be updated; the pricing card at the top of this page always shows the current rate.

Yes, OmniHuman 2 API supports multi-person animation, allowing developers to create videos featuring multiple characters in one scene. This makes it useful for virtual interviews, AI podcasts, and group performances. Results may depend on the reference image, audio input, and scene complexity.

OmniHuman 2.0 API supports longer videos of up to 300 seconds at 720p, making it ideal for extended talking avatar videos and character performances. OmniHuman 1.5 API focuses on shorter videos of up to 60 seconds, with support for 1080p resolution. Both models offer audio-driven animation, lip sync, facial expressions, and body gestures. Kie AI provides access to both OmniHuman 1.5 API and OmniHuman 2.0 API, allowing developers to choose the right model for their projects.

Yes. OmniHuman AI API supports an optional text prompt of up to 300 characters, written in Chinese, English, Japanese, Korean, Spanish (Mexican) or Indonesian, that helps guide character actions, gestures, camera movement and performance styles. Developers can combine prompts with images and audio to influence the generated animation. The final movements also depend on the audio and model interpretation.

Yes, by default. The enable_watermark parameter defaults to true; set it to false in the request input to get the video without a watermark.

Kie AI offers free credits to new users to get started with OmniHuman 2.0 API. Whether these credits cover your video generation depends on your input settings and the estimated credit cost. You can check the required credits before submitting a generation request.