← Back to blog

How to Use Gemini Omni Flash: Setup, Costs and Limits

The fastest video model on our canvas: about 35 seconds at the median for a short clip. What it costs per duration, the documented regional restriction, and when to reach for it.

Correction, 2026-10-01: this page said 1080p and 4K are upscaled "here" and listed 40-second extensions as if the canvas offered them. On 8frame, Omni Flash returns 720p clips of 3 to 10 seconds; the upscales and extensions are on Google's own API. It also told you to expect about 40 seconds (the measured median is about 35 for a 3 to 5 second clip) and ranked Kling v3 as more directable and Seedance 2.5 as better at physics, which we have not measured.

Correction, 2026-09-30: generation times are now measured over 120 days and split by model; the earlier figures lumped every Kling and every Seedance model together and printed job counts.

Gemini Omni Flash is the model to reach for when you need to try many things quickly. On our own production jobs it returns a short 720p clip (3 to 5 seconds) in about 35 seconds at the median, against about a minute for a 5-second Kling v3 Standard clip and 3 to 4 minutes for Seedance 2.0. It also generates audio by default rather than as a paid extra.

That speed is the reason to use it, and it matters more than it sounds: iteration count drives output quality more than model capability does.

TL;DR

What it costs

Canon 8frame prices, a credit is $0.01 at pack rate:

Duration Credits Approx Per second
3s 51 $0.51 17.0
4s 68 $0.68 17.0
5s 85 $0.85 17.0
8s 135 $1.35 16.9
10s 169 $1.69 16.9

Audio is generated by default at every duration, so there is no silent tier and no separate audio pass to budget for.

Comparison worth knowing: per second of finished video with audio, Omni at 16.9 and Kling v3 Standard at 17.2 are within two percent of each other. The headline numbers differ only because default clip lengths differ, 8 seconds against 5.

The documented limits

These come from Google's own model documentation and are worth knowing before you plan a workflow:

Model: gemini-omni-1.1-flash. Omni 1.1 was announced 2026-08-27.

Ratios: 16:9 default, 9:16 supported.

Resolution: native 360p and 720p. 1080p and 4K are upscaled rather than generated natively. On 8frame there is no resolution setting and clips come back at 720p.

Duration: up to 40 seconds total with extensions on Google's API. The 8frame tool makes one clip of 3 to 10 seconds per generation.

Uploaded video input is capped at 10 seconds. If your workflow starts from longer source footage, this is a hard stop.

Editing or extending uploaded video is unavailable in the EEA, Switzerland and the UK. This is a regional block, not a setting, and it is the constraint most likely to surprise a European team mid-project.

How to use it

  1. Open a canvas and add a video generation node.
  2. Pick Gemini Omni Flash from the model list.
  3. Set the ratio first. 9:16 costs the same as 16:9, and cropping later throws away most of the frame.
  4. Set the duration. Price scales linearly, so a 3-second clip at 51 credits is the cheap way to test a direction.
  5. Write the prompt with the audio in mind, since sound is generated whether you plan for it or not. Name the ambience you want or you get the model's guess.
  6. Generate. At the median, our 3 to 5 second clips came back in about 35 seconds.

When to reach for it

Exploration. The speed is the whole argument. Eight attempts in ten minutes beats two attempts in the same time, and most of the distance between a bad clip and a good one is framing and prompt rather than model capability.

Anything with speech or ambient sound. Audio arrives in the same pass, so it is synchronised rather than laid over.

Short social content. 8 seconds with audio at 135 credits is a complete social clip.

Testing a direction before committing to an expensive finish on Veo 3.1 Standard at 448.

When not to

Long source footage. The 10-second input cap rules it out.

European editing workflows. The EEA, Switzerland and UK restriction on editing uploaded video.

A silent take from a frame you already have. Omni always bills sound. Kling v3 Standard animates a start image at 57 credits for 5 silent seconds.

Long single takes. Omni stops at 10 seconds on 8frame; Seedance 2.5 and Wan 3.0 run to 30.

1080p delivery. Omni Flash returns 720p on 8frame and there is no video upscaler on the canvas, so for a 1080p deliverable pick a model that renders it, such as Veo 3.1 Fast or Wan 3.0 at 1080p.

Prompting for audio

Since sound is not optional, name it:

Unnamed audio is the model's guess, and the guess is frequently a generic music bed you did not want.

FAQ

How much does Gemini Omni Flash cost? 135 credits for 8 seconds with audio, about $1.35. 51 credits for 3 seconds.

How fast is it? About 35 seconds at the median for a 3 to 5 second clip on our production jobs; one in ten took longer than 46 seconds. Video edits took about 49 seconds. We have not measured enough 8-second clips to quote them. Times for every model are in how long does AI video generation take.

Can I feed it my own video? Up to 10 seconds. Editing or extending uploaded video is unavailable in the EEA, Switzerland and the UK.

Does it generate audio? Yes, by default, at every duration. There is no silent tier.

Sources


Explore fast here, commit elsewhere. The 8frame canvas is free and unlimited; generation is paid from $19/month.

Written with AI assistance. Corrections are noted on the page. Editorial policy

Related articles

comparisonGemini Omni Flash vs Kling v3: Speed Against ControlcomparisonGemini Omni vs Veo 3.1: Which Google ModelguideHow to Use Kling v3 Pro: When 114 Credits Pays Off

Make it
move.

Stay in the loop

Be the first to hear about our launch and get product updates