Use OmniHuman 1.5 when a person, pet, or illustrated character should perform from a finished voice track while the source portrait remains the visual anchor. Subject identification, subject detection, and mask-based selection are not available on this page.
The final cost is rounded up to whole credits.
Example output
OmniHuman 1.5
Image + Audio to Video
Example output
OmniHuman 1.5
Best for
Supports
Try this model
No results to display yet.
Model examples
Review examples to help decide whether the model fits your task.
Model strengths
Model strengths
Suggested uses
Tasks to consider
Model details
Provider
ByteDance
Category
Video
Capabilities
Image + Audio to Video
Credit charges
Credit rate: 27 per second of input
Inputs
1 image + 1 audio file
Prompt requirements
Optional prompt
Developer access
OmniHuman 1.5 can be called through Public API v1. Modes that use reference media require uploads through the Files API. Check the model reference for supported inputs.
OmniHuman 1.5 FAQ
OmniHuman 1.5 is designed for a focused image-and-audio workflow: one portrait supplies the subject, and one voice track drives the resulting performance. It fits presenter clips, character narration, and short talking-subject videos better than a general prompt-to-video task.
Upload exactly one JPEG, PNG, or WebP image up to 10 MB and one MP3, M4A, WAV, AAC, or OGG audio file up to 10 MB. The verified audio duration must be greater than zero and strictly less than 60 seconds.
Rivya uses a rate of 27 credits per verified audio second. The decimal duration remains part of the full calculation, and only the reservation total is rounded up to a whole credit. When valid actual usage no higher than the reservation is reported, Rivya settles to it; missing, invalid, or higher-than-reserved usage remains in reconciliation instead of triggering a hidden extra charge.
Fast mode prioritizes shorter processing time with a possible quality tradeoff. Leave it off for a quality-first run, and turn it on when turnaround matters more for an early test.
Yes, but it is optional and limited to 300 characters on Rivya. The current prompt path accepts Chinese, English, Japanese, Korean, Spanish, or Indonesian text; the portrait and audio remain the required inputs.
Create an audio-driven character video from exactly one portrait image and one audio clip. The subject can be a person, pet, or illustrated character, and output can be 720p or 1080p.
Choose a clear portrait that already represents the subject you want to keep, then prepare the final voice take. The audio must be shorter than 60 seconds, and both files must pass upload checks before generation.
Choose 720p or 1080p, decide whether to use the faster processing option, and leave the seed at -1 for a random result unless you need a more repeatable comparison.
The details are public, but uploading and generating require sign-in. OmniHuman 1.5 costs 27 credits per verified audio second, with only the final calculated total rounded up.
Stay here to test one portrait and one voice track. Open Studio when you need saved takes, several characters, repeated voice versions, or a longer production workflow.
Compare alternatives
Create Seedance 2.5 videos from text, a start frame with an optional end frame, or image, video, and audio references. Choose 480p, 720p, or 1080p, a duration from 4 to 30 seconds or automatic duration, and an MP4 or MOV format request.
Why consider it
Input option: Up to 50 reference files
Generate video from text, a first frame, or media references.
Why consider it
Input option: Prompt, first frame with optional last frame, or image/video/audio references
ByteDance's Seedance 2.0 Fast video model for text prompts, a first frame with an optional last frame, or image, video and audio references. Choose 480p or 720p, set a 4–15-second duration, and request audio or a returned last frame.
Why consider it
Input option: Prompt, first frame with optional last frame, or image/video/audio references
