Kling 4.0 prompt length: 8,000 tokens vs Omni's 20,000 characters

Kling 4.0 is announced with prompts up to 8,000 tokens; Sume's catalog states prompt ≤20000 characters for gemini-omni-flash-1.1. These are different units, so the two numbers do not convert one to the other. For other Sume video ids, the docs point to capabilities from GET /v1/video-router/models; no other prompt cap is quoted here.

Kling facts are from its blog post (full model "coming in October 2026"); Sume facts are from Video Router and the API catalog, read 2026-09-30.

What does Kling say about prompt length?

The post says Kling 4.0 "supports prompts up to 8,000 tokens" and can expand and refine instructions. It gives no character figure.

What is documented for Sume?

The catalog constraint for gemini-omni-flash-1.1 reads prompt ≤20000 characters. The Video Router docs describe it as one catalog id that Sume routes by the shape of the request.

Why can't I compare the two numbers directly?

A token is a model-specific chunk of text, so characters per token varies by language and tokenizer. Do not estimate a ratio from these two lines. If you are near either limit, shorten the prompt or split the story into separate clips.

What should I do about a long prompt?

Lead with subject, action, setting and camera, as in the Gemini Omni Flash prompt guide, and put detail the shot needs rather than everything you can say. Read the model's capabilities before relying on a limit that is not in the docs.