Omni references

Omni-reference models accept multiple reference types in a single generation - for example up to 9 images, 3 videos, and 3 audio clips on supporting models.

Use omni references for complex scenes: several characters, a product, a location, and a voice sample, all in one request.

Notes:

• Audio reference clips must be under 15 seconds.
• If you get a validation error with many references attached, reduce the count and retry - limits vary by model.

Updated on: 07/07/2026