Veo vs Gemini Omni 1.2 Flash: Price, Limits and What Changed in 2026

If you built on Veo 3.1 through the Gemini API, it stopped working on October 22, 2026. The model Google points you to is Gemini Omni Flash, now at version 1.2. We switched every talking clip in ppl.studio from Veo to Omni on October 12, so the notes below come from renders, not only from the spec sheet. Where a number is Google's, we say so; where it is ours, we say which model it was measured on.

The Short Version

• The latest Google video model is Gemini Omni 1.2 Flash (gemini-omni-1.2-flash), generally available since November 5, 2026.
• It costs about $0.07 a second at 720p. That is $12 per million video tokens, 31% below Omni 1.1 Flash at $17.50.
• Clips run 3 to 15 seconds, and extend to 30. Veo 3.1 makes 4, 6 or 8 seconds.
• You can edit a clip by talking to it. “Make the phone invisible. Keep everything else the same.” Veo cannot.
• Veo 3.1 is gone from the Gemini API. It lives only on Google Cloud, needs a cloud project instead of an AI Studio key, and there is no Veo 4.
• The old Omni 1.1 id turns into 1.2 on December 3, 2026, at 1.2 prices, so nobody stays on 1.1.

What Happened to Veo

Veo is not dead. Veo 3.1 is generally available on Google Cloud, and Google lists its retirement as May 4, 2027 or later. But on the Gemini API, the one a normal AI Studio key reaches, Google's own deprecation table names Omni as the replacement for each Veo model.

Specs Side by Side

What only Veo on Google Cloud still gives you is enterprise plumbing: a native negative prompt, 4K on the Standard tier, provisioned throughput, customer-managed keys and data residency. What only Omni gives you is editing, longer clips, reference images and a price per second that no Veo tier beats once the audio is on, except Veo Lite.

What Each Costs

Omni is billed in video tokens, not seconds: 720p is 5,792 tokens a second, so at $12 per million a second costs $0.0695. By resolution, 1.2 comes to about $0.023 a second at 360p, $0.104 at 1080p and $0.209 at 4K. Google upscales 1080p and 4K from the base render, so for most social video 720p is the one to pay for. Veo prices above are Google Cloud list prices; Veo is cheaper without audio, but a talking clip needs the audio.

For a four-clip, 32-second talking video that is $2.22 on Omni 1.2 against $3.23 on Omni 1.1. The only Google model cheaper per second than Omni 1.2 is Veo 3.1 Lite, and it is a preview on Google Cloud, so using it means a second sign-in path for about 28% off.

Limits That Matter

• Paid tier only. A key whose project has no billing cannot render.
• Requests, not seconds, are capped. The first paid tier allows 8 requests a minute and 40 a day (the next tier 16 and 200, then 40 and 2,000). Four clips make one video, so one key makes about ten videos a day.
• No audio in. Google says uploading audio inputs or references is unsupported, so you cannot hand Omni a voice recording to lip-sync.
• Edits on your own footage are not offered in the EEA, Switzerland or the UK. Extending and editing a video the model made works everywhere.
• Two-frame interpolation stops at 10 seconds, and 11 to 15 second clips take one reference video instead of three.
• Real people and celebrities are blocked, and every clip carries a SynthID watermark.

What We Measured

We have run talking videos on Veo 3.1 Lite and Omni 1.1 Flash, on real accounts, since early October. Nothing below is measured on 1.2 yet, and we say so in each row.

On the outside, reviews are thin and mostly from resellers. Segmind's 20 tests of Omni 1.2 found it strong on labelled diagrams, kinetic type, and background or wardrobe edits, and weaker on photoreal people from scratch and on counting past three; it reported about one instruction in four skipped. A review of the June preview reported lip-sync holding for 6 to 7 seconds before drifting, which we have not tested past 8 seconds and will, since 1.2 allows 15-second clips. Leaderboards rank Omni Flash first for text-to-video, but those rows are for 1.0 and 1.1. We found no blind head-to-head of Omni 1.2 against Veo 3.1.

Prompting Omni

• Ask for one continuous shot. By default Omni cuts into a few shots and invents a story. Say “single continuous shot, no scene cuts”.
• Put exclusions in the prompt. There is no negative field, so write “No dialogue. No extra sound effects.” after the subject and action.
• Audio is part of the prompt. Left alone it invents a soundtrack. Ask for the sound you want or for none.
• Be specific about motion from a still. “Slow push-in, handheld wobble, she nods” beats “make it move”. Use a sharp first frame.
• Time it. “After 3 seconds, she lifts the bottle” or a block like [0-5s] … [5-10s] … works.
• Keep edits short. “Make the phone invisible. Keep everything else the same.” Long edit prompts change other things.
• Quote any text you want on screen, word for word.
• The duration field wins over the prompt. A length written in the text is overridden by the request's duration, and over 15 seconds is an error.

Which One to Use

What This Means for ppl.studio

Talking workflows such as the creator UGC video render on Gemini Omni Flash with the Gemini key you add in Settings, and silent ones such as the drama song ad default to MiniMax H3 when your account has a fal key. Google moves every Omni 1.1 request to 1.2 on December 3, so those workflows get the lower price and the longer clips from then on. We will re-run our talking-clip tests on 1.2, including 12 to 15 second takes, and update this page with the numbers. The still frames for each clip come from Nano Banana 2.1.

Frequently Asked Questions

Two of our stills, four seconds into Gemini Omni 1.2

Our own test of the new model: the same first frame and prompt as a MiniMax H3 clip, rendered on Omni 1.2 Flash.