Gemini Omni Flash 1.1 edit mode: the default reserve is 8 s, $1.00
Editing a clip with Gemini Omni Flash 1.1 on Sume reserves 8 seconds by default: $1.00 at 720p. Send a duration hint of 6 and the reserve drops to $0.75.

If you edit a clip with gemini-omni-flash-1.1 and a video_url, Sume reserves 8 seconds of output by default, which is $1.00 at 720p ($0.125 per second x 8). Add a duration hint of 6 for a six-second source and the reserve becomes $0.75. The edit price follows output seconds, so the hint is the lever you control.
The reason is in how the provider works. The edit queue on fal takes no duration, and the source clip sets the output length. Sume's docs state that fal does not return an output duration, so the reserve is a hint and not a probe, and they flag that as needing verification. Budget with the hint, and compare the settled amount against it once the first job finishes.
Edit reserve by resolution
| Resolution | Sume per second | 6 s hint | 8 s default | 10 s hint |
|---|---|---|---|---|
| 360p | $0.0375 | $0.23 | $0.30 | $0.38 |
| 720p | $0.125 | $0.75 | $1.00 | $1.25 |
| 1080p | $0.1875 | $1.13 | $1.50 | $1.88 |
| 4K | $0.375 | $2.25 | $3.00 | $3.75 |
The arithmetic
Provider list per second on the row is $0.03 at 360p, $0.10 at 720p, $0.15 at 1080p and $0.30 at 4K, each multiplied by 1.25. At 720p: $0.10 x 1.25 = $0.125, and $0.125 x 8 = $1.00. At 4K: $0.30 x 1.25 = $0.375, and 8 seconds is $3.00. With the 5.5% platform fee, the 720p default is about $1.06.
Edit requests send no aspect_ratio and no duration to the provider, and a request that includes either an aspect_ratio or a generate_audio: false is a 400 error. The model always produces native synced audio, so there is no cheaper silent variant.
Steps
- Probe the source length, then send
durationas that length rounded up (maximum 10). - Choose the resolution: 720p is the default and the cheapest that keeps detail for social.
- Send
video_urlalone. It cannot be combined withimage_url,end_image_urlorreference_*_urls. - Compare the
usage.coston the finished job with the table; the source clip sets the real output length.
A worked batch
Say a team edits 40 product clips, each 6 seconds, replacing one object per clip. With the duration hint set to 6 at 720p the reserve is 40 x $0.75 = $30.00. Leave the hint off and every job reserves 8 seconds, so the reserve is 40 x $1.00 = $40.00, a $10.00 difference in held balance even if the settled amounts end up close.
At 1080p the same batch reserves 40 x 6 x $0.1875 = $45.00 with the hint. At 360p it is 40 x 6 x $0.0375 = $9.00, which is useful for checking that an edit prompt does what you intend before you pay for sharper pixels. Add the 5.5% fee from Sume's pricing policy to any of these to get the all-in figure.
What Sume does not do
Sume does not edit audio references, does not expose extend, and does not route sume/auto into edit unless you send a video_url; an edit with no key or a failed decision falls back to this same Omni 1.1 row. Numbers are the repo's constants read on 2026-10-08.
Sources
Related posts
More in Pricing
- generation_spend_cap_usd of 0, 120, 450, null or 600: which applies?
On a Format run 120 and 450 are used as given, null is the $500 platform maximum, and 0 or 600 returns a 400. Omitted means the Format's cap or $400.
- GLM-5.3 off-peak 50% points vs a Sume dollar spend cap
Z.ai meters GLM-5.3 in points with a 50% off-peak rate. Why a Sume dollar cap is a different meter, and how to schedule agent work.
- GPT Image 2.5 on fal runs $0.00402 to $0.40026 per image: on Sume
fal's ChatGPT Image 2.5 Flare page lists $0.00402 for 1024x768 low and $0.40026 for 3840x2160 max. At Sume's 1.25 multiplier that is about $0.005 to $0.50.
- Does gpt-image-2.5 cost more at 4K? Prices from 1024 to 3840x2160
gpt-image-2.5 price by pixel size on Sume: 1024x1024, 1536x1024, 2048x2048 and 3840x2160 at medium and high, plus the custom-size rules from the docs.
Written by Sume