comparisons

Gemini Omni 1.2 Flash vs Kling 4.0 Flash vs Veo 3.1

Poyo.ai Team
18 min read
Share:

Key takeaways

  • Three fast video models, three different strengths. Gemini Omni 1.2 Flash gives the most flexibility (3–15 s, 360p to 4K, editing). Kling 4.0 Flash makes the longest single clips (up to 20 s) and the widest set of aspect ratios at 720p. Veo 3.1 is the cheapest way on PoYo to get a finished 8-second clip with audio.
  • Gemini Omni vs Veo is now a succession question. Google shut down the Veo 3.1 preview models in the Gemini API on October 22, 2026 and named gemini-omni-1.2-flash as their replacement. Veo 3.1 stays available as a GA model on Google Cloud, and both remain available on PoYo.
  • Clip length decides a lot. Veo 3.1 generates 4, 6 or 8 seconds per request. Omni 1.2 Flash takes any whole second from 3 to 15. Kling 4.0 Flash takes any whole second from 3 to 20.
  • Per clip on PoYo: a 5-second 720p clip costs $0.41 on Omni 1.2 Flash and $0.47 on Kling 4.0 Flash. An 8-second Veo 3.1 clip with audio costs $0.10 (Lite) or $0.18 (Fast) on the fixed-length Veo 3.1 page, and $0.60 on Veo 3.1 Fast Official.
  • One key for all three on PoYo. The same API key, balance and submit/status endpoints run every model here, on pay-as-you-go credits with no subscription.
  • Editing existing footage: Omni 1.2 Flash and Kling 4.0 Flash both edit a source video with a text instruction. Veo 3.1 Official extends a clip with 1–7 extra seconds.

Gemini Omni 1.2 Flash vs Kling 4.0 Flash vs Veo 3.1 at a glance

Gemini Omni 1.2 FlashKling 4.0 FlashVeo 3.1
DeveloperGoogleKuaishou (Kling AI)Google
PoYo model IDgemini-omni-1.2-flashkling-4.0-flashveo3.1-lite / -fast / -quality, or veo3.1-lite-official / -fast-official / -quality-official
Clip length3–15 s, any whole second (up to 10 s with a last frame)3–20 s, any whole second4, 6 or 8 s (Veo 3.1 page: fixed 8 s)
Resolution360p, 720p, 1080p, 4K720p720p, 1080p, 4K (by tier)
Aspect ratios16:9, 9:1616:9, 9:16, 1:1, 21:916:9, 9:16
Text-to-videoYesYesYes
Image-to-videoFirst frame, or first + last frameOne starting frameOne image, or first + last frame
Reference inputs on PoYoUp to 10 images + 3 videosUp to 10 images, 5 videos, 7 subjects (15 total)3 reference images (Fast and Quality, 8-s clips)
Edit an existing videoYes, up to 15 sYes, 3–20 s, with reference videosExtension on Veo 3.1 Official
PoYo priceFrom $0.0273/s (360p); $0.0819/s at 720p$0.094/s; $0.063 per billed second with video inputFrom $0.10 per 8-s video, or from $0.018/s on Official

All three models run through the same PoYo endpoint, so you can test them side by side with one API key: Gemini Omni 1.2 Flash, Kling 4.0 Flash, Veo 3.1 and Veo 3.1 Official.

Gemini Omni vs Veo: which versions are you comparing?

Searches for "Gemini Omni vs Veo" started when Google first released Gemini Omni Flash, and many of the pages that rank still compare the early preview with Veo 3.1. Google has since shipped two more Omni versions and changed where Veo 3.1 is offered:

ModelGoogle model IDReleasedStatus on November 14, 2026
Veo 3.1 and Veo 3.1 Fast (preview)veo-3.1-generate-preview, veo-3.1-fast-generate-previewOctober 15, 2025Shut down in the Gemini API on October 22, 2026; replacement: gemini-omni-1.2-flash
Veo 3.1 and Veo 3.1 Fast (GA, Google Cloud)veo-3.1-generate-001, veo-3.1-fast-generate-001November 17, 2025Generally available on Gemini Enterprise Agent Platform; retirement May 4, 2027 or later
Veo 3.1 Lite (preview)veo-3.1-lite-generate-previewMarch 31, 2026Shut down in the Gemini API on October 22, 2026
Gemini Omni Flash (preview)gemini-omni-flash-previewJune 30, 2026Shut down on October 22, 2026
Gemini Omni 1.1 Flashgemini-omni-1.1-flashAugust 27, 2026From December 3, 2026, requests route to 1.2
Gemini Omni 1.2 Flashgemini-omni-1.2-flashNovember 5, 2026Current version

Two practical points follow from this table:

  • In Google's own developer API, Omni has taken over from Veo. If you build on the Gemini API directly, Omni 1.2 Flash is the video model to use. Veo 3.1 continues on Google Cloud for enterprise customers.
  • Older "Omni Flash vs Veo 3.1 Fast" comparisons describe the preview. Omni 1.2 Flash makes longer clips (15 s, up from 10 s on 1.1), edits full 15-second clips, and Google describes it as faster and more cost-effective. For a full look at the previous version, see our Gemini Omni 1.1 Flash guide.

The rest of this article compares the current models you can call today: Gemini Omni 1.2 Flash, Kling 4.0 Flash, and the Veo 3.1 Lite, Fast and Quality tiers.

What the three models share

  • All three accept text and images. Each one turns a prompt into a clip and animates a starting image.
  • All three are positioned for speed. Kling calls Flash "significantly faster generation at exceptional value", Google describes Omni 1.2 Flash as faster and more cost-effective, and Veo 3.1 Fast is the speed tier of the Veo family. Generation time grows with clip length, resolution and current load, so time a few of your own prompts on each model.
  • Google's two models generate sound with the picture. Omni 1.2 Flash produces native audio, and Veo 3.1 generates sound too; on Veo 3.1 Official you choose audio or a silent clip with the sound field.
  • All three use the same asynchronous flow on PoYo. Submit a task, get a task_id, then poll the status endpoint or receive a callback.

Specs compared: length, resolution, aspect ratio and inputs

Clip length

ShortestLongest per requestStepsDefault on PoYo
Gemini Omni 1.2 Flash3 s15 s (10 s with a last frame)1 s8 s
Kling 4.0 Flash3 s20 s1 s5 s
Veo 3.1 Official4 s8 s4, 6, 8 sSelected per request
Veo 3.18 s8 sFixed8 s

For a 12-, 15- or 20-second shot in one take, the two Flash models cover it in a single request. Veo 3.1 Official reaches longer lengths through veo3.1-extend, which adds 1–7 seconds to an existing clip, up to a 30-second total.

A single 15-second Gemini Omni 1.2 Flash clip: a four-shot macro explainer inside a beehive. Example from the PoYo model page.

Resolution and framing

  • Gemini Omni 1.2 Flash generates at 360p or 720p, and Google upscales to 1080p or 4K. That makes 360p a cheap draft tier and 4K a delivery tier inside the same model. Clips are 16:9 or 9:16.
  • Kling 4.0 Flash outputs 720p in 8-bit SDR, according to Kling. It adds 1:1 and 21:9 framing, which suits square feed posts and ultrawide banners. When you animate a starting image, the image sets the frame.
  • Veo 3.1 offers 720p, 1080p and 4K in 16:9 or 9:16 at 24 FPS. On PoYo, 4K is available on Fast and Quality, and on the fixed-length Lite tier. Veo 3.1 Lite Official offers 1080p on 8-second clips.

An 8-second 16:9 text-to-video clip from Veo 3.1 Fast Official: a street interview beside yellow taxis in New York. Example from the PoYo model page.

Inputs and editing

WorkflowGemini Omni 1.2 FlashKling 4.0 FlashVeo 3.1 on PoYo
Text promptpromptpromptprompt
Start frameimage_urls with one imageimage_urls with one imageOne image (Fast, Quality, Official tiers)
First and last frameimage_urls with two images, up to 10 sStarting frame (end frames on Kling 4.0)Two images
Reference imagesUp to 10Up to 103 (Fast, Quality; 8-s clips)
Reference videosUp to 3, each up to 15 sUp to 5, total input video up to 20 sImage references
Reusable subjectsReference images and videosUp to 7 kling_elements (2–4 images or one 3–8 s video each)Image references
Edit a source videovideo_url up to 15 s; output follows the source lengthvideo_url 3–20 s, plus reference images, up to 4 reference videos and subjectsExtension with veo3.1-extend

Kling 4.0 Flash has the richest reference system: you can name each input in the prompt as @Image1, @Video1 or @Element1, and reuse a subject across many clips. Omni 1.2 Flash is the simplest to edit with: send the clip and one sentence describing the change, and the edit follows the source timing. Veo 3.1 is the most direct for frame-to-frame transitions and short finished shots.

Kling 4.0 Flash video editing: a corgi bounding through snow (left) re-rendered as a blocky voxel corgi with the same motion (right). Example from the PoYo model page.

Pricing: per second and per clip on PoYo

All prices below are PoYo list prices ($1 = 200 credits). One PoYo API key covers all three models, billing is pay-as-you-go with no subscription, and you pay per second or per clip as listed.

Per-second rates

ModelRate on PoYo
Gemini Omni 1.2 Flash360p $0.0273/s · 720p $0.0819/s · 1080p $0.1229/s · 4K $0.2457/s
Kling 4.0 Flash720p $0.094/s for text, image and image-reference jobs · $0.063 per billed second when video is an input
Veo 3.1 Official, Lite720p $0.018/s silent, $0.03/s with audio · 1080p (8-s clips) $0.03/s silent, $0.048/s with audio
Veo 3.1 Official, Fast720p/1080p $0.05/s silent, $0.075/s with audio · 4K $0.15/s silent, $0.175/s with audio
Veo 3.1 Official, Quality720p/1080p $0.12/s silent, $0.24/s with audio · 4K $0.24/s silent, $0.36/s with audio
Veo 3.1 (fixed 8 s, audio included)Lite $0.10 · Fast $0.18 · Quality $1.00 per video at 720p/1080p; 4K $0.15 / $0.30 / $2.00

Price per clip at 720p: the Flash models

ClipGemini Omni 1.2 FlashKling 4.0 Flash
4 s$0.33$0.38
5 s$0.41$0.47
8 s$0.66$0.75
10 s$0.82$0.94
15 s$1.23$1.41

Kling 4.0 Flash continues up to 20 seconds for $1.88. Exact values: Omni 1.2 Flash 5 s = 81.9 credits ($0.4095), 10 s = 163.8 credits ($0.819). Kling 4.0 Flash 5 s = 94 credits ($0.47), 10 s = 188 credits ($0.94).

Price per clip at 720p: Veo 3.1

ClipVeo 3.1 Official Fast, with audioVeo 3.1 Official Lite, with audio
4 s$0.30$0.12
6 s$0.45$0.18
8 s$0.60$0.24

An 8-second Veo 3.1 Official Fast clip with audio is 120 credits ($0.60). On the fixed-length Veo 3.1 page, an 8-second clip with audio is $0.18 on Fast and $0.10 on Lite.

Worked examples

JobGemini Omni 1.2 FlashKling 4.0 FlashVeo 3.1
Ad test: 30 vertical 8-s variants at 720p30×0.6552 = $19.6630×0.752 = $22.56Fast, fixed 8 s: 30×0.18 = $5.40
Draft then deliver: ten 8-s drafts + one 8-s 4K final360p drafts 10×0.2184 + 4K 1.9656 = $4.15Ten 720p drafts 10×0.752 = $7.52 (720p is the final)Lite Official drafts 10×0.144 + Quality Official 4K with audio 8×0.36 = $4.32

Editing and video references on the Flash models

JobGemini Omni 1.2 FlashKling 4.0 Flash
Edit a 10-s clip10×0.0819 = $0.82Billed 10+10 = 20 s × $0.063 = $1.26
8-s clip guided by one 5-s reference video8×0.0819 = $0.66Billed 8+max(5,8) = 16 s × $0.063 = $1.01

Omni 1.2 Flash bills the output seconds, with reference videos and editing included in that rate. Kling 4.0 Flash uses a lower per-second rate with video input and bills the output duration plus the larger of the input video length and the output duration. For the full Kling billing rules, see our Kling 4.0 API guide.

Kling 4.0 Flash vs Gemini Omni 1.2 Flash: head to head

These two are the closest match: both are "Flash" tiers, both cover text, image, reference and editing modes behind one model ID, and both bill per second.

  • Choose Kling 4.0 Flash for 16–20-second clips, square or ultrawide framing, recurring characters defined as subjects, and reference videos that set motion or camera paths. Its references count against a 15-item budget, and input videos together can run up to 20 seconds.
  • Choose Gemini Omni 1.2 Flash for resolution range (360p drafts to 4K delivery), first-and-last-frame transitions, simple prompt-based editing, and reference-heavy jobs where video inputs stay inside the per-second rate.
  • On price at 720p, Omni 1.2 Flash is about 13% cheaper per second ($0.0819 vs $0.094) for text and image jobs, and cheaper again once reference videos or editing are involved.

If you are choosing inside the Kling family, read Kling 4.0 vs Flash for the differences in resolution, clip length and frame control. For a full walkthrough of the Flash request fields, see the Kling 4.0 Flash API guide.

Where Seedance 2.5 fits

"Gemini Omni vs Seedance" comes up often as well. ByteDance's Seedance 2.5 sits in a different tier: it targets longer, reference-heavy production clips of 4–30 seconds at 480p, 720p or 1080p, with up to 50 image, video and audio references and optional synchronized audio. On PoYo a 5-second 720p Seedance 2.5 clip from text or images costs $1.575 (63 credits per second), about four times the Omni 1.2 Flash price. Use it for the final cut of a complex scene, and keep the Flash models for high-volume drafts. For a visual side-by-side, see Kling 4.0 vs Seedance 2.5.

Which should you choose?

Your jobStart withWhy
Cheapest finished 8-s clip with audioVeo 3.1 Lite (veo3.1-lite) for text prompts, Veo 3.1 Fast (veo3.1-fast) with images$0.10 or $0.18 per video at 720p or 1080p
Silent B-roll at the lowest per-second costVeo 3.1 Lite Official, sound: false$0.018/s at 720p
Clips longer than 8 s in one takeGemini Omni 1.2 Flash (up to 15 s) or Kling 4.0 Flash (up to 20 s)One request covers the whole shot
Square posts or 21:9 bannersKling 4.0 FlashOffers 1:1 and 21:9 framing
Cheap drafts, then 4K delivery from the same modelGemini Omni 1.2 Flash360p at $0.0273/s, 4K at $0.2457/s
Recurring characters and products across a seriesKling 4.0 FlashUp to 7 reusable subjects, 15 references
Restyle or change the setting of an existing clipGemini Omni 1.2 FlashEdits up to 15 s, billed by the source length
Edit a clip and bring in a new character or motionKling 4.0 FlashEditing also accepts reference videos and subjects
Planned start-to-end transitionGemini Omni 1.2 Flash or Veo 3.1Both take a first and a last frame
Highest Veo fidelity for a hero shotVeo 3.1 Quality4K with audio on the Official tier

A practical routing setup: draft concepts on Omni 1.2 Flash at 360p or on Veo 3.1 Fast, move approved ideas to the model that matches the delivery format, and track cost per accepted clip.

How to run all three on PoYo

PoYo gives you all three models through one API key and one request format, with pay-as-you-go credits and no subscription. Every model uses the same submit and status endpoints. This request runs Gemini Omni 1.2 Flash at 720p:

curl https://api.poyo.ai/api/generate/submit \
  -H "Authorization: Bearer $POYO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gemini-omni-1.2-flash",
    "input": {
      "prompt": "A barista slides a latte across a sunlit counter. Slow push-in, warm morning light, soft cafe ambience.",
      "duration": 10,
      "aspect_ratio": "9:16",
      "resolution": "720p"
    }
  }'

The same body works for Kling 4.0 Flash: change "model" to "kling-4.0-flash". Both accept duration 10, aspect_ratio 9:16 and resolution 720p.

For Veo 3.1 Official, pick a tier, a duration of 4, 6 or 8 seconds, and the sound setting:

curl https://api.poyo.ai/api/generate/submit \
  -H "Authorization: Bearer $POYO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "veo3.1-fast-official",
    "input": {
      "prompt": "A barista slides a latte across a sunlit counter. Slow push-in, warm morning light, soft cafe ambience.",
      "duration": 8,
      "aspect_ratio": "9:16",
      "resolution": "720p",
      "sound": true
    }
  }'

Each submission returns data.task_id. Check progress with:

curl https://api.poyo.ai/api/generate/status/YOUR_TASK_ID \
  -H "Authorization: Bearer $POYO_API_KEY"

The status moves through not_started, running, and finished or failed. When it is finished, the video URL is in data.files[].file_url. You can also add a top-level callback_url to receive a webhook when the video is ready. Each model page has a Playground with a live credit estimate and a JSON view you can copy into your code.

FAQ

Is Gemini Omni better than Veo?

They aim at different jobs. Gemini Omni 1.2 Flash is the more flexible model: 3–15-second clips, 360p to 4K, multiple references and prompt-based editing. Veo 3.1 makes short 4–8-second shots and has a Quality tier for higher-fidelity output. Google now offers Omni 1.2 Flash as the replacement for Veo 3.1 in the Gemini API, while Veo 3.1 continues as a GA model on Google Cloud.

Is Gemini Omni Flash replacing Veo 3.1?

In the Gemini API, yes. Google shut down veo-3.1-generate-preview, veo-3.1-fast-generate-preview and veo-3.1-lite-generate-preview on October 22, 2026 and lists gemini-omni-1.2-flash as the replacement. The GA models veo-3.1-generate-001 and veo-3.1-fast-generate-001 remain on Gemini Enterprise Agent Platform and are scheduled to stay available until at least May 4, 2027. On PoYo, both Veo 3.1 pages remain available.

Gemini Omni Flash vs Veo 3.1 Fast: which is cheaper?

It depends on clip length. For an 8-second clip with audio at 720p on PoYo, Veo 3.1 Fast costs $0.18 on the fixed-length Veo 3.1 page and $0.60 on Veo 3.1 Official, against $0.66 for Omni 1.2 Flash. For shorter clips or 360p drafts, Omni 1.2 Flash can cost less: a 5-second 360p draft is $0.14. For clips longer than 8 seconds, Omni does it in one request.

How long can each model's videos be?

Gemini Omni 1.2 Flash: 3–15 seconds (up to 10 with a first and last frame). Kling 4.0 Flash: 3–20 seconds. Veo 3.1: 4, 6 or 8 seconds per request, with extension on Veo 3.1 Official up to a 30-second total.

Which fast video model supports 4K?

Gemini Omni 1.2 Flash (upscaled 4K, $0.2457/s on PoYo) and Veo 3.1 (4K on Fast and Quality, and on the fixed-length Lite tier). Kling 4.0 Flash outputs 720p; for 1080p in the Kling family, use Kling 4.0.

Gemini Omni vs Kling: which is better for consistent characters?

Kling 4.0 Flash gives you more structure: up to 7 reusable subjects, each defined by 2–4 images or a short video, plus up to 10 reference images and 5 reference videos in one request. Omni 1.2 Flash accepts up to 10 reference images and 3 reference videos and handles character swaps through editing. For series work with the same cast, start with Kling 4.0 Flash.

Is Gemini Omni free?

Google includes 15-second Omni clips in Google Vids at no additional cost with a Google account or Workspace subscription, and in the Gemini app for Google AI Ultra subscribers. For API access, PoYo offers Omni 1.2 Flash on pay-as-you-go credits from $0.0273 per second at 360p, with no subscription, and the Playground shows the credit estimate before each run.

Can I use one API key for all three models?

Yes. On PoYo, Gemini Omni 1.2 Flash, Kling 4.0 Flash and every Veo 3.1 tier share one API key, one balance and the same submit and status endpoints. Omni and Kling even accept the same request body for a text prompt, so switching between them is a change to the model field.

Share: