AI Video Models by ByteDance · Rp1,131.92 / second audio · about 30 to 600 seconds · available now
Turn a photo and audio into a 720p avatar video with motion and lip sync, lasting 4–300 seconds.
Capabilities
720p
4–300 seconds
Lip sync
Sample output
Sample output from OmniHuman 2.0, made on GetRun.
Input parameters
Name
Type
Required
prompt
string
no
audio_url
string
yes
image_urls
array
yes
enable_watermark
boolean
no
The complete schema is published as JSON Schema at GET https://getrun.id/v1/models.
How to call it through the API
Try OmniHuman 2.0 in Studio, or call it from code: POST https://getrun.id/v1/tasks with body {"model": "omnihuman-2-0", "input": {…}} and an Idempotency-Key header, then poll GET https://getrun.id/v1/tasks/{task_id}. The details are in the model documentation.
OmniHuman 2.0 is priced at Rp1,131.92 / second audio and billed from a rupiah balance as you use it, with no required subscription. The estimated cost is shown before it runs, and failed tasks are never charged.
How do I call OmniHuman 2.0 through the API?
Send POST https://getrun.id/v1/tasks with an Authorization: Bearer header and an Idempotency-Key, body {"model": "omnihuman-2-0", "input": {...}}, then poll GET https://getrun.id/v1/tasks/{task_id} every 2 to 5 seconds until the status is succeeded.
How long does OmniHuman 2.0 take per request?
OmniHuman 2.0 typically takes 30 to 600 seconds, depending on load and the input parameters.
Are OmniHuman 2.0 results stored permanently?
No. Output URLs are served from a CDN and expire after 7 days, so copy the files to your own storage if you need them longer.