Reference limits: Veo 3.1, Gemini Omni Flash and Sume Video 1.0
Veo 3.1 takes up to 3 reference images, Omni Flash up to 3 clips of 3 s each, and Sume Video 1.0 takes 1 to 9 images. A table to pick by what you need to pin.

Veo 3.1 takes up to 3 reference images, Gemini Omni Flash takes up to 3 reference clips of 3 seconds each, and Sume Video 1.0 takes 1 to 9 reference images. Pick by what you need to keep fixed: faces and products fit image references, motion and framing fit clip references.
| Model | Image references | Video references | Audio references |
|---|---|---|---|
| Veo 3.1 (Gemini API) | Up to 3 (needs 8 s) | Not listed in the source | Not listed in the source |
| Gemini Omni Flash (Gemini API) | Not stated in the source | Up to 3 clips, each up to 3 s | Unsupported |
| Sume Video 1.0 | 1 to 9 | Not listed | Not listed |
What each source says
The Veo guide lists up to 3 reference images and notes that reference images require an 8 second clip. The Omni guide lists a maximum of 3 reference clips of up to 3 seconds each, says audio references are unsupported, and says negative prompts are not available. We read both on 2026-10-03; where a cell says not listed, the page we read did not state it.
Sume's side
Sume documents these by model, and the lists are machine readable. supported_input_references at GET /v1/videos/models tells you which of image_url, video_url and audio_url a model accepts. The docs list Video 1.0, now a retiring compatibility alias for Video Router Auto, with 1 to 9 reference_image_urls, and gemini-omni-flash-1.1 with image and video references but no audio reference, and video edits through video_url on the Video Router. See the Sume video docs.
A quick rule
- Need a product or a face to stay the same across shots: use image references and check the cap first.
- Need a camera move or a motion to carry over: use a video reference and keep it under the clip limit.
- Need to match a voice: none of the listed models above takes an audio reference, so plan a separate audio step.
Sources
Related posts
More in Comparisons
- Video API price per second: Veo, Runway, Luma, MiniMax, Seedance
A vendor-authored Higgsfield table lists Veo 3.1 at $0.40 per second down to MiniMax H3 at $0.08. What 10 seconds costs and why to verify each rate.
- Sume vs Argil: AI avatar video and video agents compared
Argil makes AI-avatar and story videos with a chat agent, Director; Sume is a video agent with a multi-model API. Avatars, API, pricing, and limits compared.
- Sume vs fal: a generative media API or a video agent platform
fal runs 1,000+ image, video, and audio models behind one API. Sume adds a video agent, Formats, and avatars to a multi-model API. How the two surfaces differ.
- HeyGen alternatives with an API: price units, limits, and fit
HeyGen alternatives with an API: Synthesia, Creatify, Argil, Arcads, and Sume compared by price unit, API shape, limits, and live vs rendered avatars.
Written by Sume