Pika Video Studio's 50 references vs Sume's per-model limits
Pika Video Studio attaches up to 50 references. Sume's caps are per model: 10 images and 3 short clips on Omni Flash, 1 to 8 images on Genjutsu.

Pika's new Video Studio lets you attach up to 50 characters, products, settings and styles and @mention them in a prompt. Sume has no single 50-item library. Its limits belong to each model: Gemini Omni Flash 1.1 takes up to 10 reference images and up to 3 reference videos of 3 seconds or less, and the Genjutsu motion-transfer model takes 1 to 8 reference images.
Pika's number is from its September 17 post. Sume's are from Video Router, read 2026-10-01.
How does Pika describe its references?
The post says Video Studio attaches up to 50 characters, products, settings and styles, which you then @mention in the prompt to make a clip or a multi-shot video up to 30 seconds with sound. It does not say how many items each underlying model receives.
What are Sume's reference limits?
Limits are set per catalog model, so read capabilities from GET /v1/video-router/models instead of assuming one envelope.
| Model | Reference inputs |
|---|---|
gemini-omni-flash-1.1 | reference_image_urls up to 10, reference_video_urls up to 3, each 3 seconds or less |
| Genjutsu (Motion Transfer) | One source video_url plus 1 to 8 reference_image_urls |
| Seedance 2.x, Wan 3.0, MiniMax H3 | Audio and video references honored, per the Video generation page |
How do I point at one reference in a prompt?
On Omni Flash, address media as <IMAGE_REF_0> and <VIDEO_REF_0>. The numbers are 0-based and follow list order, so the first image you send is <IMAGE_REF_0>. That is Sume's counterpart to an @mention, with positions instead of names. The Seedance form is covered in Seedance 2.5 @ tags.
What if I have more than 10 references?
Choose the ones that decide the shot. If a character appears in many images, send the clearest few. Splitting a long story into several shots, each with its own small reference set, also stays inside the caps. Send an Idempotency-Key on each request so a retry does not create a duplicate job.
Sources
Related posts
More in Models
- Qwen-Audio-3.0-TTS languages: 16, and Sume's language field
Qwen-Audio-3.0-TTS lists 16 languages. On Sume TTS you set the language field to a BCP-47 or ISO-639 code for every non-English transcript.
- Qwen-Image 2.1 ControlNet Union Fun in ComfyUI vs reference images
ComfyUI added Qwen-Image 2.1 ControlNet Union Fun for editing on 2026-09-29. Sume has no ControlNet input; it edits with reference images and masks.
- Qwen Image 2.1 license: research only, commercial needs a request
Qwen-Image-2.1 ships under the Qwen Research License: non-commercial means research or evaluation only, and commercial use needs a separate license.
- Recraft V4 Styles: 10 references, and the Sume Recraft id
Recraft V4 Styles takes one to 10 reference images in Precise or Flexible mode. Sume lists recraft/recraft-v4 as text-to-image only and rejects references.
Written by Sume