Veo 3.1 Lite 1,024-token prompt limit vs Omni's 20,000 characters

The Gemini API's Veo 3.1 Lite model page lists a text input limit of 1,024 tokens, with text and image input and video with audio output (read 2026-10-04). That is room for a tight single-shot prompt, not a full script with timing marks.

Sume does not list Veo 3.1 Lite, but it does list a Google video model: gemini-omni-flash-1.1. Its catalog row sets a prompt limit of 20,000 characters, 3 to 10 second clips, ratios 16:9 and 9:16, and resolutions up to 4K. Tokens and characters are different units, so the two numbers are not a strict ratio, but the Omni row leaves far more space for detail.

When prompt length matters

If your prompt is short, the limit never bites and price decides: the Gemini pricing page, read 2026-10-04, lists Veo 3.1 Lite at $0.05 per second at 720p and $0.08 at 1080p, while Sume bills Omni 1080p at $0.1875 per second.

The Gemini API pricing page is the source for the Veo numbers.

• Shot-by-shot descriptions with camera moves for each second of a 10-second clip.
• Brand rules, character descriptions and negative instructions pasted in with the shot.
• Dialogue lines with timing, when audio is generated with the clip.

Send a long prompt

The video generation docs show the body. This request puts a multi-line prompt on Omni.

import os
import requests

prompt = "\n".join([
    "0-3s: wide shot of a night market, handheld, warm neon.",
    "3-6s: push in on a vendor flipping skewers, steam rising.",
    "6-10s: close on a customer smiling, shallow focus.",
])
body = {
    "model": "gemini-omni-flash-1.1",
    "prompt": prompt,
    "duration": 10,
    "resolution": "1080p",
    "aspect_ratio": "9:16",
}
r = requests.post(
    "https://api.sume.com/v1/videos",
    headers={
        "Authorization": f"Bearer {os.environ['SUME_API_KEY']}",
        "Idempotency-Key": "omni-long-001",
    },
    json=body,
)
print(r.status_code)