Gemini Omni 1.1 is billed on tokens, not on flat per-clip prices. Your bill for a request is:
(video tokens x $21.875/M) + (text output tokens x $11.25/M) + (input tokens x $1.875/M)
Video tokens are the dominant term, and the model emits a fixed number of them per second of output, set only by resolution:
| Resolution | Video tokens per second | Video cost per second |
|---|---|---|
| 360p | 1,931 | $0.0422/s |
| 720p | 5,792 | $0.1267/s |
| 1080p | 8,688 | $0.1900/s |
| 4k | 17,376 | $0.3801/s |
Because that count is fixed, cost is linear in duration (3–10 seconds) and the per-second rate is exact, not an estimate. 1080p is 1.5x the 720p rate, 4k is 3x, and 360p is one third.
What "draft" means on 360p. It is a label about intended use, not a reduced model. 360p runs on the same endpoint with the same feature set: text-to-video, first/last frame, reference images and the same synchronized audio track all work exactly as they do at higher resolutions. The only differences are the frame size, 640x360 instead of 1280x720, and the price, one third of 720p. It is called a draft tier because that price makes it the right place to iterate on prompt, timing, composition and sound before you commit to a delivery render.
On top of the video tokens, each request adds a small text-output charge for the model's own reasoning, typically 250–750 tokens, or about $0.003 to $0.008 per request. Your prompt and any input images are billed at $1.875 per million and are usually well under a cent.
What You Actually Pay
Typical all-in cost, including the reasoning overhead.
| Resolution | 3s | 5s | 8s | 10s |
|---|---|---|---|---|
| 360p | $0.13 | $0.22 | $0.34 | $0.43 |
| 720p | $0.39 | $0.64 | $1.02 | $1.27 |
| 1080p | $0.58 | $0.96 | $1.53 | $1.91 |
| 4k | $1.15 | $1.91 | $3.05 | $3.81 |
Audio
Every Gemini Omni 1.1 clip is delivered as an MP4 containing both a video track and a synchronized stereo 48 kHz AAC audio track. There is no audio parameter on this model: audio is always generated and cannot be turned off.
There is also no separate audio charge. The per-second video token counts above already cover the audio track, and no audio token line appears on the bill. Writing detailed sound design into your prompt costs the same as leaving audio unspecified, so describe the sound you want.
Important Notes
- •Duration scales exactly linearly — a 10s clip costs precisely 2x a 5s clip at the same resolution, because video tokens per second are fixed.
- •Aspect ratio is free — 16:9 and 9:16 produce identical token counts and identical bills.
- •Resolution is the only significant cost lever — everything else moves the total by cents.
- •Reference media is nearly free — a first/last frame, or up to 10 reference images, only adds input tokens, typically a fraction of a cent and under $0.02 even at 10 images.
- •The seed is free — fixing a seed for reproducibility costs nothing.
Cost Optimization Tips
- •Iterate at 360p, deliver at 720p or above. 360p costs one third of 720p, so ten drafts plus one 720p final costs about the same as five 720p attempts.
- •Fix the seed before you scale up. Set an explicit seed on the 360p take you approved so the higher-resolution render follows it, instead of paying 4k rates to re-roll a composition.
- •Match resolution to the destination. 720p covers social and web delivery at a third of the 4k rate. Reserve 4k for broadcast, large-format or VFX plates.
- •Trim duration to the shot. You pay per second, so a tight 5s cut costs half a 10s one. Use bracketed timecodes in the prompt to fit the action into the shorter clip.
Example Scenario
A team producing 30 vertical clips a day at 9:16, 720p, 5 seconds spends about $19.2/day ($0.64 x 30), or roughly $575/month. Drafting each clip at 360p first and rendering only the approved takes at 720p brings that closer to $250/month for the same output volume.