Veo 3.1 prompts cap at 1,024 tokens; Sume's Omni at 20,000 characters

Google's Veo 3.1 page, read 2026-10-01, gives a text input limit of 1,024 tokens. Sume does not list Veo. Its Google video row is gemini-omni-flash-1.1, whose catalog constraint is "prompt ≤20000 characters".

The two limits use different units, so they do not convert exactly. A token is a piece of text that is usually shorter than a word, so count Veo prompts against the token limit and Omni prompts against the character limit.

What fits in a prompt that short?

A one-shot video prompt rarely needs more than subject, action, camera, lighting and sound. If you are hitting a cap, move fixed detail out of the text: send the character or product as a reference image instead of describing it. On Sume, gemini-omni-flash-1.1 takes up to 10 reference images and up to 3 reference videos of 3 seconds each, and addresses them in the prompt as <IMAGE_REF_0> and <VIDEO_REF_0>, in list order.

What does the call look like?

Reference tokens are 0-based and the reference media is sent before the prompt. One reference image with no first or end frame makes the request reference-to-video. See the video docs for the field list.

curl -X POST https://api.sume.com/v1/video-router/generate \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: omni-ref-001" \
  -d '{
    "model": "gemini-omni-flash-1.1",
    "prompt": "<IMAGE_REF_0> walks through a night market, handheld, neon reflections",
    "reference_image_urls": ["https://example.com/character.png"],
    "resolution": "720p",
    "aspect_ratio": "16:9",
    "duration": 6
  }'

How do I check the limit before I submit?

Count characters on your side against the 20,000 cap for Omni. Other Sume models list their own constraints on the model catalog; read the one for the model you pick instead of assuming the Omni number carries over.

How do the two prompt limits line up?

Limits from the sections above.