Omni Reference to Video API
Use the same service in your application: upload references, pass their file IDs by type, select a model and duration, and poll the generated task.
Read the Omni Reference to Video API referenceAdd reference media, describe the shot, and create a video with Wan 3.0. Wan 3.0 accepts up to 10 reference images, up to 5 reference videos, and up to 5 audio clips, with a combined reference duration of at most 15 seconds. Check the credit estimate before you start.
Start with 3,000 free trial credits for eligible tasks. Compare free credits and usage limits.
Wan 3.0 accepts up to 10 reference images, up to 5 reference videos, and up to 5 audio clips, with a combined reference duration of at most 15 seconds. Only the inputs the selected model supports are shown.
Model: Loading model…
| Model | Credit rate |
|---|
Rates are loaded for your account. For a rate billed per second, credits = rate × video duration. The total below is estimated by Medux for your selected model, duration, clarity, and aspect ratio.
Sign in to see the current credit cost for each model.
The latest task appears below. Processing continues on the server if you leave this page.
No video yet. Generate one to see it here.
Use the service behind this Workshop in your application or AI agent. Explore the API reference and its corresponding MCP tool before starting an integration.
Use the same service in your application: upload references, pass their file IDs by type, select a model and duration, and poll the generated task.
Read the Omni Reference to Video API referenceThe corresponding medux_video_generate_omni_reference_to_video tool lets a connected agent create a video from reference media.
Open the Omni Reference MCP toolUpload the reference images, and a clip where the model accepts one. Wan 3.0 accepts up to 10 reference images, up to 5 reference videos, and up to 5 audio clips, with a combined reference duration of at most 15 seconds.
Tell Wan 3.0 how the references should appear, then select a supported duration, clarity, and aspect ratio.
Check the estimate against your balance before starting, then preview or download the completed video.
Keep a character or product recognizable across a new shot by supplying it as a reference.
Use reference images, or a short clip where supported, to steer look and movement.
Combine several references in one video instead of relying on a single starting image.
Say in the prompt what each reference is for, such as the character, the setting, or the camera motion.
Wan 3.0 accepts up to 10 reference images, up to 5 reference videos, and up to 5 audio clips, with a combined reference duration of at most 15 seconds.
The quote uses your account, duration, clarity, and aspect ratio. Final usage appears with the finished task.
Every Medux tool has open examples on GitHub: copy a prompt for Claude or Codex, or start from a working API request.
A ready-to-paste prompt and approval checklist for running medux_video_generate_omni_reference_to_video from Claude.
View on GitHubThe same workflow for Codex: expected tool call, arguments to review and how to verify the result.
View on GitHubRequest fields, curl, Python and JavaScript samples, and response handling for this endpoint.
View on GitHubWan 3.0 accepts up to 10 reference images, up to 5 reference videos, and up to 5 audio clips, with a combined reference duration of at most 15 seconds.
Sign in to see the current rate for each model. Medux estimates the total from the selected model, duration, clarity, and aspect ratio before you submit. The completed task shows its final cost.
This page accepts one reference video and one audio clip. Use the Playground or the API when you need several.
Yes. Once the service returns a task ID, processing continues on the server. Return with the same account to restore your latest task, or find it in Account Results.