Skip to content
Model infoModelsSnapshot: 15 Oct 2026

ByteDance / OmniHuman

OmniHuman 1.5

Image-and-audio character video generation with optional text direction and subject masks. BytePlus, Volcengine and fal expose different limits and prices.

Video generation
availableClosed weightsVideo generation
Official documentation (opens in a new tab)Reviewed 15 Oct 2026
7 of 21 specification fields collected
Output resolution720p, 1080pSee specification notes below
Clip durationNot collectedAwaiting source review
Input billingSee ratesfal
Output billingSee ratesPublished units and conditions below

Specifications

Provider-documented facts. “Not collected” means we have not verified that field yet, not that the feature is unsupported.

Context / input limit
Not collected
Maximum input tokens
Not collected
Maximum output
Not collected
Reasoning
Not collected
Tool calling
Not collected
Structured output
Not collected
Streaming
Not collected
Batch processing
Not collected
Fine-tuning
Not collected
Knowledge cutoff
Not collected
Release date
Not collected
Parameters
Not collected
Architecture
Multimodal language-model guidance combined with a Multimodal Diffusion Transformer and Pseudo Last Frame designSource for Architecture (opens in a new tab)
Languages
Not collected
Context caching
Not collected
Output resolutions
720p, 1080pSource for Output resolutions (opens in a new tab)

Published hosted output options; default 1080p. Audio duration differs by host.

Output formats
MP4Source for Output formats (opens in a new tab)
Input modalities
Image, Audio, TextSource for Input modalities (opens in a new tab)
Output modalities
VideoSource for Output modalities (opens in a new tab)
Published training data
15,000 hours of filtered video; poor lip/audio alignment caused audio dropout for 70% of samples. Final quality tuning used the top 100 hours.Source for Published training data (opens in a new tab)
Published research configuration
Paper experiments: 120 frames at 24 fps and 480p, then separate super-resolution to 720p/1080p; longer clips generated autoregressively.Source for Published research configuration (opens in a new tab)

Research configuration, not an API frame-rate or clip-length guarantee.

Access & pricing

Prices belong to an endpoint and billing tier. A provider listing does not guarantee access for every account or region.

fal

Provider source (opens in a new tab)
API model ID
fal-ai/bytedance/omnihuman/v1.5
Access
Documented hosted API; service activation, authentication, billing and account eligibility required. No paid inference was run.
Regions
Not collected
Supported input workflows
One image plus one audio fileSource for Supported input workflows (opens in a new tab)
Audio duration limits
Strictly under 30 seconds at 1080p; under 60 seconds at 720pSource for Audio duration limits (opens in a new tab)
Generation mode
Asynchronous queued task with status polling and webhook supportSource for Generation mode (opens in a new tab)
Billing basis
Returned duration measures audio input / generated video duration for billing.Source for Billing basis (opens in a new tab)
Output formats
Video file URL; exact container is not constrained by the generic output schema.Source for Output formats (opens in a new tab)
Output resolutions
720p, 1080pSource for Output resolutions (opens in a new tab)
Default resolution
1080pSource for Default resolution (opens in a new tab)
Prompt controls
Optional text directionSource for Prompt controls (opens in a new tab)
Mask behavior
Optional mask_url string; the subject in the white region speaks.Source for Mask behavior (opens in a new tab)
Generation controls
turbo_mode defaults to false; faster generation trades off quality.Source for Generation controls (opens in a new tab)

Prices retain their published currencies and billing units below.

Published billing units · original currency · observed 15 Oct 2026
CategoryUsageRateUnit
Video generationGenerated video duration1$0.16second

Rate conditions

  1. Published host-specific video-duration price. Account access, applicable taxes, rounding and contract conditions may affect the final bill. No token conversion or text-question estimate is inferred.
Pricing source (opens in a new tab)

Data & enterprise terms

Retention, training use, regional processing and enterprise terms have not been summarized for this endpoint.

BytePlus Vision AI

Provider source (opens in a new tab)
API model ID
realman_avatar_picture_omni15_cv
Access
Documented hosted API; service activation, authentication, billing and account eligibility required. No paid inference was run.
Regions
ap-singapore-1
API endpoint URL
https://cv.byteplusapi.comSource for API endpoint URL (opens in a new tab)
Hosted API version
2024-06-06Source for Hosted API version (opens in a new tab)

API protocol version, not model release date.

Generation mode
CVSubmitTask / CVGetResult; cancellation while in_queue or generatingSource for Generation mode (opens in a new tab)
Authentication
HMAC-SHA256 signed request with access key and secretSource for Authentication (opens in a new tab)
Audio duration limits
Strictly under 60 secondsSource for Audio duration limits (opens in a new tab)
Image input limits
Below 5 MB; resolution below 4096 × 4096Source for Image input limits (opens in a new tab)
Result and task expiry
Video URL valid for one hour; tasks may expire after 12 hours.Source for Result and task expiry (opens in a new tab)

URL/task expiry does not establish data-retention or training policy.

Output resolutions
720p, 1080pSource for Output resolutions (opens in a new tab)
Default resolution
1080pSource for Default resolution (opens in a new tab)
Mask behavior
Optional mask_url string; multiple subject-mask URLs separated by semicolons.Source for Mask behavior (opens in a new tab)
Seed control
Default -1 (random); fixed positive seed can improve reproducibility.Source for Seed control (opens in a new tab)
Generation controls
pe_fast_mode defaults to false; trades quality for speed.Source for Generation controls (opens in a new tab)
Published request quota
One concurrent generationSource for Published request quota (opens in a new tab)
Languages
Chinese, English, Japanese, Korean, Mexican Spanish, IndonesianSource for Languages (opens in a new tab)

Text prompt languages, not a speech-recognition or TTS language list.

Prices retain their published currencies and billing units below.

Published billing units · original currency · observed 15 Oct 2026
CategoryUsageRateUnit
Video generationGenerated video duration1$0.12second

Rate conditions

  1. Published host-specific video-duration price. Account access, applicable taxes, rounding and contract conditions may affect the final bill. No token conversion or text-question estimate is inferred.
Pricing source (opens in a new tab)

Data & enterprise terms

Retention, training use, regional processing and enterprise terms have not been summarized for this endpoint.

Volcengine Jimeng AI

Provider source (opens in a new tab)
API model ID
jimeng_realman_avatar_picture_omni15_v2
Access
Documented hosted API; service activation, authentication, billing and account eligibility required. No paid inference was run.
Regions
cn-north-1
API endpoint URL
https://visual.volcengineapi.comSource for API endpoint URL (opens in a new tab)
Hosted API version
2022-08-31Source for Hosted API version (opens in a new tab)

API protocol version; the selector suffix v2 does not mean OmniHuman 2.0.

Audio duration limits
Strictly under 60 seconds; under 15 seconds recommended for qualitySource for Audio duration limits (opens in a new tab)
Image input limits
Below 5 MB; resolution below 4096 × 4096Source for Image input limits (opens in a new tab)
Output resolutions
720p, 1080pSource for Output resolutions (opens in a new tab)
Default resolution
1080pSource for Default resolution (opens in a new tab)
Output formats
MP4Source for Output formats (opens in a new tab)
Mask behavior
Optional array of subject-mask URL strings, unlike BytePlus’s semicolon-separated string.Source for Mask behavior (opens in a new tab)
Seed control
Default -1 (random); fixed positive seed can improve reproducibility.Source for Seed control (opens in a new tab)
Prompt length limit
300 charactersSource for Prompt length limit (opens in a new tab)

Overview states a maximum; API reference describes 300 characters as recommended.

Generation controls
pe_fast_mode defaults to false. To match Jimeng, provider recommends true at 720p and false at 1080p.Source for Generation controls (opens in a new tab)
Generation mode
Asynchronous CVSubmitTask / CVGetResultSource for Generation mode (opens in a new tab)
Published request quota
One concurrent generationSource for Published request quota (opens in a new tab)
Result and task expiry
Video URL valid for one hour; task may expire after 12 hours.Source for Result and task expiry (opens in a new tab)
Content marking controls
Optional AIGC metadata on retrieval; response reports whether marking succeeded.Source for Content marking controls (opens in a new tab)

Prices retain their published currencies and billing units below.

Published billing units · original currency · observed 15 Oct 2026
CategoryUsageRateUnit
Video generationGenerated video duration1CN¥1.00second

Rate conditions

  1. Published host-specific video-duration price. Account access, applicable taxes, rounding and contract conditions may affect the final bill. No token conversion or text-question estimate is inferred.
Pricing source (opens in a new tab)

Data & enterprise terms

Retention, training use, regional processing and enterprise terms have not been summarized for this endpoint.

Caching

Caching can reuse earlier input processing. Availability, lifetime and billing depend on the endpoint. A cache hit still generates a new response.

01 / INITIAL INPUTProcess & retain

The system processes eligible input and retains reusable state.

02 / LATER REQUESTCheck available state

Reuse depends on matching input, session state and the cache policy.

03 / CACHE HITReuse & generate

Reusing state can reduce repeated work. Any charges follow the billing rules below.

fal

fal-ai/bytedance/omnihuman/v1.5
Not verified
How it works
Not collected
Minimum reusable input
Not collected
Lifetime & refresh
Not collected
How to enable it
Not collected
Usage fields
Not collected
Billing rules
Not collected
Limits & data boundaries
Not collected

No verified cache rate for this endpoint. An empty rate is not a zero charge.

Caching reviewed 15 Oct 2026

BytePlus Vision AI

realman_avatar_picture_omni15_cv
Not verified
How it works
Not collected
Minimum reusable input
Not collected
Lifetime & refresh
Not collected
How to enable it
Not collected
Usage fields
Not collected
Billing rules
Not collected
Limits & data boundaries
Not collected

No verified cache rate for this endpoint. An empty rate is not a zero charge.

Caching reviewed 15 Oct 2026

Volcengine Jimeng AI

jimeng_realman_avatar_picture_omni15_v2
Not verified
How it works
Not collected
Minimum reusable input
Not collected
Lifetime & refresh
Not collected
How to enable it
Not collected
Usage fields
Not collected
Billing rules
Not collected
Limits & data boundaries
Not collected

No verified cache rate for this endpoint. An empty rate is not a zero charge.

Caching reviewed 15 Oct 2026

What the measurements say

No benchmark observations have been added for this model. We do not turn a vendor description into a performance score.

Latency and throughput depend on the host, request size, reasoning settings and load. There is no measured speed or global usage figure in this record.

The record behind the model

Why it is included

Inferred from the named ByteDance proprietary hosted offering, original research project and reviewed ByteDance/ByteDance-Seed public weight inventories, which contain no matching OmniHuman checkpoint. The original project states that downloads are not offered. This is not a published weight-license declaration.

Classification evidence (opens in a new tab)

Lifecycle dates

Original research paper first submitted26 Aug 2025Source (opens in a new tab)

Sources reviewed