All three are priced by the second. Two of them, Grok and Omni, tell you how long a clip can be. Veo 3.1, Gemini Omni Flash and Grok Imagine Video are the current video models from Google and xAI, and they are documented so differently that a straight comparison is mostly a map of what each company will tell you.
- Grok Imagine Video is the most documented — 1 to 15 seconds, 480p by default, 1080p available.xAI, Video Generation, read at source 16 Sep 2026: “The allowed range is 1–15 seconds”.
- Veo 3.1 has the widest price range and the sharpest top end — from $0.05 to $0.60 a second, and of these three it is the one with a published 4K price. Its cheapest tier undercuts both rivals; its 4K tier costs several times more than either rival’s published rate.First-hand: until 22 Sep 2026 this line said Veo was “the one with a 4K tier”. Gemini Omni Flash also outputs 4K; see the next line.Google, Gemini Developer API pricing, read in a browser 16 Sep 2026 and re-read 17 Sep 2026: Veo 3.1 Standard “$0.60 (4k)”; Veo 3.1 Lite “$0.05 (720p)”.
- Gemini Omni Flash is the odd one out. It is an editing model as much as a generator. Google gives its clips as 3 to 10 seconds, at 360p up to 4K, but prices it only per second of 720p.Google, Gemini Omni Flash model page, last updated 27 Aug 2026, read in a browser 22 Sep 2026: output video “3s-10s (360p/720p/1080p/4K, 24 FPS)”.First-hand: until 22 Sep 2026 this line said Google “publishes no duration figure” for Omni. The model page had carried one since 27 Aug 2026, before this page was written on 16 Sep, so the page was wrong, not out of date.
- Omni has a free route the others do not — Google says it is rolling out “at no cost to users on YouTube Shorts and YouTube Create App”.Google, Introducing Gemini Omni, 19 May 2026, read at source 16 Sep 2026.
- On the API pages read here, only Google says its output is watermarked. Omni carries SynthID. xAI’s FAQ for the Grok website and apps says Grok output is watermarked too: “Generated images and videos include a Grok watermark … There is no setting to remove the watermark.”xAI, FAQ – Grok Website / Apps, last updated 27 Aug 2026, read at source 23 Sep 2026: “Generated images and videos include a Grok watermark to indicate that the content was created with AI. There is no setting to remove the watermark.”First-hand: until 23 Sep 2026 this line said “Only Google says its output is watermarked.” xAI’s Grok FAQ, last updated 27 Aug 2026, already said Grok output carries a watermark when this page was written on 16 Sep, so the line was wrong, not out of date. The FAQ covers the Grok website and apps; the API pages read here do not mention a watermark.
Every price here was correct when this site last checked it, between 16 Sep and 22 Sep 2026. Prices change often and this site no longer updates them, so check the vendor’s own page before you rely on one: Google pricing · xAI pricing · Black Forest Labs pricing · BytePlus pricing.
What each vendor publishes
| Veo 3.1 | Gemini Omni Flash | Grok Imagine Video 1.5 | |
|---|---|---|---|
| Vendor | xAI | ||
| Clip length | not published | 3–10 seconds | 1–15 seconds |
| Top resolution | 4K | 4K (1080p and 4K by upscaling) | 1080p |
| Default resolution | not published | 720p | 480p |
| API price | $0.05–$0.60 a second | about $0.10 a second at 720p | $0.08 a second |
| Free route | none | YouTube Shorts | none published |
| Audio | included in the price | yes, with the video | on by default |
| Watermark | not stated here | SynthID | not stated on the API pages; the Grok app FAQ says a Grok watermark |
Google, Gemini Developer API pricing, read in a browser 16 Sep 2026 · Google, Introducing Gemini Omni, 19 May 2026 · xAI, Video Generation and pricing, both read at source 16 Sep 2026 · Google, Gemini Omni Flash model page and Gemini API changelog, read in a browser 22 Sep 2026: “Resolution control: New resolution parameter in video_config supports 360p, 720p (default), 1080p, and 4k outputs. 1080p and 4K outputs are generated using upscaling.” · xAI, FAQ – Grok Website / Apps, read at source 23 Sep 2026: “Generated images and videos include a Grok watermark”.
The blanks in this table are the finding. Google has filled Omni’s: a clip runs 3 to 10 seconds, so you can now price an Omni clip at 720p, not just an Omni second. What an Omni second costs above 720p is still not on the pricing page.First-hand: until 22 Sep 2026 this takeaway said Omni’s clip length was unpublished. Google’s model page gave it from 27 Aug 2026.
What Gemini Omni actually is
Not a straight text-to-video model. Google describes it as a model “that can create anything from any input — starting with video”, and the demonstrations are edits: “Take a video you shot and just ask Omni to change what’s happening.”Google, Introducing Gemini Omni, 19 May 2026, read at source 16 Sep 2026.
The pitch is that it holds a scene across turns: “Refine your videos across multiple turns. Change the environment, angle, style or even specific details, without ever losing the thread of your original scene.” Google also claims a physics improvement — “an improved intuitive understanding of forces like gravity, kinetic energy and fluid dynamics”.
It is on the API now, and priced by the second like Veo. Google’s pricing page lists Gemini Omni Flash as “now generally available to developers on the paid tier of the Gemini API”, with billing “based on total output token consumption, calculated at a rate of 5,792 tokens per second of 720p video”, which it puts at “approximately $0.10 per second”.Google, Gemini Developer API pricing, read in a browser 16 Sep 2026 (the docs refuse automated fetches). Per-token rates: input “$1.50 (text / image / video / audio)”, output “$17.50 (video)”.
The generally available model, gemini-omni-1.1-flash, arrived on 27 Aug 2026 with a resolution setting: 360p, 720p by default, 1080p or 4K, the top two made by upscaling. Clips run 3 to 10 seconds, and a video you bring in for editing or extension can be up to 10 seconds. The older preview endpoint, gemini-omni-flash-preview, is scheduled to shut down on 30 Sep 2026.Google, Gemini API changelog, 27 Aug 2026 entry, read in a browser 22 Sep 2026: “The existing gemini-omni-flash-preview endpoint will be deprecated on September 30, 2026.” · Gemini Omni Flash model page, read in a browser 22 Sep 2026: input “Video (up to 10s for editing and extension)” · Gemini API deprecations, read in a browser 22 Sep 2026: gemini-omni-flash-preview, shutdown “September 30, 2026”.
Where to use it: “Gemini Omni Flash is rolling out today to all Google AI Plus, Pro and Ultra subscribers globally through the Gemini app and Google Flow”, and “at no cost to users on YouTube Shorts and YouTube Create App”. Flow is the surface that absorbed Whisk and ImageFX — the explainer is here.
Every Omni output is marked. Google: “All videos created with Omni include our imperceptible SynthID digital watermark.”
What Grok Imagine Video gives you instead
Numbers. xAI documents duration, resolution, aspect ratio and defaults, and the defaults are low — 480p unless you ask, 16:9 unless you ask. Editing an existing clip keeps its length, “capped at 8.7 seconds”, and reference-to-video tops out at 720p. The full control surface is on the Grok Imagine cheat sheet.
And Veo 3.1
The widest price range here, and the only one with a published 4K price. Omni also outputs 4K, by upscaling, but Google gives its rate only at 720p. Google prices it by the second rather than the clip and lists no free tier; the arithmetic for a single shot is on what Veo actually costs, and the prompt shape on the Veo cheat sheet.
Two more video models priced by the second, for scale. Black Forest Labs, known for FLUX image models, put FLUX 3 out as a preview on 4 Aug 2026: “FLUX 3 is our first video model: one model trained across image, video, and audio, generating video with synchronized sound.” Text- or image-to-video costs $0.17 a second at hd, $0.29 at fhd, $0.40 at qhd and $0.80 at uhd, or $0.06 a second for a draft.Black Forest Labs, Release notes, 4 Aug 2026 entry, read at source 22 Sep 2026: “Available now as a preview.” · API Pricing, read at source 22 Sep 2026: “Pay per image, or per second for video”. ByteDance bills Seedance 2.5 by token on BytePlus: $10.70 per million tokens for 480p and 720p output and $11.7 for 1080p, without a video input. BytePlus works that out, for a 5-second 16:9 clip, at $0.103 a second at 480p, $0.231 at 720p and $0.569 at 1080p.ByteDance, BytePlus ModelArk Pricing, page last updated 22 Sep 2026, read in a browser 22 Sep 2026: dreamina-seedance-2-5-260628 “For 480p and 720p outputs: Input without video: 10.70”, and for 1080p “11.7” (USD per million tokens). Re-query before relying on these; this page has not checked them since.
What this page could not verify
- Output quality. Nothing here was generated or compared side by side.
- What an Omni second costs above 720p. The pricing page gives Omni’s rate per second of 720p only.First-hand: until 22 Sep 2026 this item said Omni’s clip length could not be found. Google’s model page gives 3 to 10 seconds.
- Whether Veo output is watermarked. Not stated on the pages read here.
Google, Introducing Gemini Omni, 19 May 2026 · Gemini Omni Flash model card · Gemini Developer API pricing · xAI, Video Generation and pricing. All read at source on 16 September 2026. Added 22 September 2026: Google, Gemini Omni Flash model page, Gemini API changelog and deprecations · Black Forest Labs, Release notes and API Pricing · BytePlus, ModelArk Pricing. Added 23 September 2026: xAI, FAQ – Grok Website / Apps.
Video model specifications move monthly.