Seedance 2.5 Flash vs Gemini Omni 1.2 Flash (2026)
We may earn a commission if you buy through links on this page. Read the affiliate disclosure · How we review tools
Head to head
Gemini Omni 1.2 Flash$0.07/sQuick answer. Use Seedance 2.5 Flash when the clip must run past 15 seconds, up to 30, or when you need documented video and audio references. Use Gemini Omni 1.2 Flash for 1080p or 4K and Google's listed $0.07 per second at 720p. Veo 3.1 is the older Google model. Official docs only.
| Tool | Rating | Best for | Price | Website |
|---|---|---|---|---|
| Seedance 2.5 Flash | 4.4/5 | Clips past 15 seconds | $0.15/s at 720p | Visit Seedance 2.5 Flash → |
| 4.4/5 | 1080p, 4K, and 720p price | $0.07/s at 720p | Visit Gemini Omni 1.2 Flash → |
For a search on seedance vs gemini omni, pick the constraint that is actually in the docs. Seedance 2.5 Flash, announced Nov 12, 2026, is the model for a clip that has to run past 15 seconds, up to 30, and for a public set of video and audio references. Gemini Omni 1.2 Flash, model id gemini-omni-1.2-flash, went generally available on Nov 5, 2026. It is the model Google’s posts price at $0.07 per second for 720p, with 360p through 4K on the Cloud model page.
Veo 3.1 is the older Google video model, and it is being phased out. The specs below are Seedance 2.5 Flash and Gemini Omni 1.2 Flash.
This page is based on official docs, pricing pages, and the vendors’ own announcements, checked Nov 17, 2026. The 4.4 marks mean each model has a documented job here. They are tied on purpose. They record a fit for a job, and they are separate from any quality score.
Flash’s own price card, the model id, and the China notes are in Seedance 2.5 Flash. Launch dates sit on the AI model release timeline. Best AI models points here from the video section.
How this comparison was made
We read BytePlus’s Dreamina Seedance 2.5 series doc and the ModelArk pricing page updated Nov 16, 2026, the same pages behind the Seedance 2.5 Flash article. For Gemini Omni 1.2 Flash we read Google’s Nov 5, 2026 announcement, the Google Cloud model page (last updated 2026-11-17 UTC), the Google Cloud general-availability post from Nov 5, 2026, and the DeepMind model card published Nov 5, 2026.
Where those pages disagree, both wordings stay. The announcement says the price is 30 percent below Gemini Omni 1.1. The Cloud blog says 30 percent less than Gemini Omni 1.1 Flash, with similar savings at other resolutions, and it does not print those other rates. The model card lists audio files as an input. The Cloud model page marks the audio modality as not supported. The Cloud blog says audio references are allowlisted. This page does not collapse those lines into one audio cap.
Where a post says “better reference following,” “near-flagship quality,” or “2x speed,” the sentence stays attached to the vendor. We did not generate a clip with either model. Flagship Seedance 2.5 appears in the price section because it is the 1080p and 4K option in the same BytePlus family. It is a sibling, and the header comparison is Flash against Omni.
Side-by-side specs
| Seedance 2.5 Flash | Gemini Omni 1.2 Flash | |
|---|---|---|
| Released | Announced Nov 12, 2026. API id dreamina-seedance-2-5-flash-261115. Pricing page updated Nov 16, 2026. | GA Nov 5, 2026. Model id gemini-omni-1.2-flash. The Cloud page lists retirement of support on that platform as Nov 5, 2027 or later. |
| API | BytePlus ModelArk. OpenRouter lists bytedance/seedance-2.5-flash. | Interactions API is supported. GenerateContent is marked not supported. Count Tokens is also listed as supported. |
| Length | 4 to 30 seconds, or automatic within that range. | Generation and editing: 3 to 15 seconds. First and last frame: 3 to 10 seconds. The announcement says one-second steps, up from 10 seconds on Omni 1.1. |
| Extension | Edit and extend are in the series doc. The doc does not print a separate extension length. | Extension output is 3 to 10 seconds, concatenated onto up to the last 30 seconds of the input. The model uses up to the last 10 seconds of an input of any duration. |
| Resolution | 480p and 720p. 1080p and 4K are not supported on this id. | 360p, 720p, 1080p, and 4K. The announcement describes drafting at 360p or 720p and upscaling to 1080p or 4K. |
| Audio output | Generate video with audio is supported. | Video output with synchronized audio. Sound generation lists speech, music, and sound effects. |
| Aspect ratio | 21:9, 16:9, 4:3, 1:1, 3:4, 9:16, adaptive. | 16:9 and 9:16. |
| References | 10 images, 5 videos, and 5 audios. 20 files in total. | Up to 10 images per prompt and up to 3 videos per prompt. See the reference section for the audio wording. |
| Editing | Edit and extend. | Video editing, extension, and upscaling. The announcement describes multi-turn conversational editing on a full 15-second clip. |
| Provenance | The BytePlus pages used here do not print a C2PA or SynthID line. | Content Credentials (C2PA) are supported on the Cloud page. The announcement, the Cloud blog, and the model card describe a SynthID watermark. |
Pricing
The invoices use different units. BytePlus publishes a token rate and 5 second examples. Google’s Nov 5 posts publish a per-second rate for 720p. A shared 5 second receipt would require a Google example those posts do not print, so this page compares the per-second figures they do print.
Seedance, from the BytePlus examples
16:9, 5 seconds, no video input. Read Nov 17, 2026, on a page updated Nov 16, 2026.
| 480p | 720p | 1080p | 4K | |
|---|---|---|---|---|
| Flash token rate | $7.0 per million tokens | same rate covers 720p | Not supported | Not supported |
| Flash example | $0.34, or $0.07 per second | $0.76, or $0.15 per second | Not supported | Not supported |
| Seedance 2.5 example | $0.514, or $0.103 per second | $1.156, or $0.231 per second | $2.843, or $0.569 per second | $4.568, or $0.914 per second |
With video input, Flash’s 5 second output example is $0.37 to $1.45 at 480p and $0.84 to $3.25 at 720p, depending on 2 to 30 seconds of input. The full card, including the $4.3 second rate and the OpenRouter listing, is on Seedance 2.5 Flash.
Gemini Omni 1.2 Flash
Google’s announcement lists $0.07 per second for 720p, and says that price is 30 percent below Gemini Omni 1.1. The Cloud blog lists $0.07 per second of 720p video, 30 percent less than Gemini Omni 1.1 Flash, with similar savings at other resolutions. Neither post prints a 360p, 1080p, or 4K rate. This page does not fill those cells. The 30 percent line is Google’s comparison with its previous Omni model. It is a vendor claim about Omni 1.1, and it is separate from the Seedance example.
On the published 720p per-second figures, $0.07 is lower than BytePlus’s $0.15 example. Seedance’s 480p example is also $0.07 per second. That match is across two resolutions. At the same 720p label, the printed figures are $0.15 and $0.07.
Use-case verdicts
Clips longer than 15 seconds
Seedance 2.5 Flash is the documented choice for one request. The series doc sets 4 to 30 seconds. Omni’s generation and editing duration is 3 to 15 seconds, and a first-and-last-frame shot is 3 to 10 seconds. A 15 second clip fits both: it is Omni’s documented maximum, and it sits inside Seedance’s range. A 20 or 30 second generation in one request is the Seedance cell.
Omni can continue a clip. Extension adds 3 to 10 seconds onto up to the last 30 seconds of the input video, using up to the last 10 seconds of that input. That is a follow-up step. It is a different job from Seedance’s single 4 to 30 second request.
1080p and 4K
Gemini Omni 1.2 Flash is the winner between these two when the file has to be 1080p or 4K. Flash lists 480p and 720p. The pricing page marks 1080p and 4K as not supported. Omni lists 360p, 720p, 1080p, and 4K. Google’s posts describe drafting at 360p or 720p and upscaling a keeper to 1080p or 4K. They do not print the upscale rate.
Flagship Seedance 2.5, id dreamina-seedance-2-5-260628, lists 1080p and 4K on the same BytePlus pages, at the example prices above, and it keeps the 4 to 30 second range. Use Omni for a 1080p or 4K shot on Google’s API. Use flagship Seedance 2.5 when that higher-resolution shot also has to run past 15 seconds from the BytePlus API.
720p price
Gemini Omni 1.2 Flash has the lower published 720p per-second figure: $0.07 against BytePlus’s $0.15 example for a 16:9 clip with no video input. Inside the Seedance family, Flash is still the cheap draft versus flagship Seedance 2.5, at $0.15 against $0.231 at 720p on that same example. The $0.15 figure is the no-input case. A video-input 720p example can run from $0.84 to $3.25 for 5 seconds of output.
Google’s 30 percent reduction is the cut versus Omni 1.1. Carry it as Google’s claim. The Cloud blog’s “similar savings at other resolutions” does not include a second dollar rate.
Reference inputs
Both docs list a cap of 10 images. Seedance also lists 5 videos and 5 audios, 20 files in total. Omni’s Cloud page lists up to 3 videos per prompt. The announcement describes reference motion clips of up to three seconds, and the spec table lists a maximum video length of 15 seconds. This page quotes both and does not pick a third number.
Audio is the messy cell. Seedance’s series doc lists 5 reference audios. Omni’s model card lists audio files among the inputs. The Cloud blog says the model follows image, video, and audio references, and then notes that audio references are gated behind allowlisting. The Cloud model page marks the audio modality as not supported, and lists sound generation as an output capability. A public Seedance audio slot and an allowlisted Omni audio reference are different access stories. This page does not state an Omni audio-file count.
Google’s announcement says Omni 1.2 Flash follows references more closely than before, and it shows that claim against Omni 1.1. BytePlus says Flash stays near flagship Seedance 2.5. Both lines are vendor descriptions. This page does not rank frames.
Aspect ratio
Seedance 2.5 Flash is the documented choice when the frame is something other than 16:9 or 9:16. Its list is 21:9, 16:9, 4:3, 1:1, 3:4, 9:16, and adaptive. Omni’s Cloud page lists 16:9 and 9:16. The announcement says a conversational edit can switch the aspect ratio. It does not add a third ratio to the spec table, so the ratios in the table stay 16:9 and 9:16.
Editing
Gemini Omni 1.2 Flash is the model whose docs name multi-turn conversational editing, video editing, extension, and upscaling. The announcement describes changing a product, a background, lighting, weather, or aspect ratio on the same clip, including a full 15-second clip. Seedance’s series doc lists edit and extend, and it does not describe a conversational edit loop or an upscale step. There is no shared quality score for edits.
Provenance
Omni’s Cloud page lists Content Credentials (C2PA) as supported. The announcement, the Cloud blog, and the model card say outputs carry a SynthID watermark. The BytePlus pages used for Flash do not print either line. That is a gap in the pages we read. It is a missing cell, and this page does not turn it into a claim that Seedance files are unlabeled.
Speed
BytePlus’s Flash model page says about 2x speed against flagship Seedance 2.5, at near-flagship quality. The Omni pages read for this comparison do not print a latency number. There is no speed winner between Flash and Omni. The 2x line is BytePlus’s claim against its own flagship.
Seedance 2.5 Flash
👍 Pros
- 4 to 30 second clips in one request, past Omni's 15 second generation cap
- 10 images, 5 videos, and 5 audios of reference on the public doc
- Aspect ratios include 21:9, 4:3, 1:1, and 3:4, as well as 16:9 and 9:16
👎 Cons
- No 1080p or 4K on this id
- The 720p example is $0.15 per second, above Google's listed $0.07
- The 2x speed line is BytePlus's claim against flagship Seedance 2.5
Gemini Omni 1.2 Flash
👍 Pros
- 360p, 720p, 1080p, and 4K, with synchronized audio on the Cloud page
- A listed 720p rate of $0.07 per second on the Nov 5, 2026 posts
- Up to 10 images and 3 videos per prompt, plus editing, extension, upscaling, and C2PA
👎 Cons
- One generation is 3 to 15 seconds, and first and last frame is 3 to 10
- The spec table lists 16:9 and 9:16
- Audio references are allowlisted on the Cloud blog, while the model page marks audio input as not supported
Who should use which
Use Seedance 2.5 Flash when the cut has to run past 15 seconds in one request, when the frame is outside 16:9 and 9:16, or when the references include video files beyond three or audio files on a public doc. Move a finished 1080p or 4K shot to flagship Seedance 2.5 if you are already on that API and the clip also has to stay in the 4 to 30 second range. The walkthrough is Seedance 2.5 Flash.
Use Gemini Omni 1.2 Flash when the file has to be 1080p or 4K, when you are spending the published $0.07 per second 720p rate, or when the job is a multi-turn edit, an extension, an upscale, or a file with C2PA. Budget 3 to 15 seconds for one generation. Read the Cloud console before you treat “similar savings at other resolutions” as a 4K price. This page does not supply one.
Veo 3.1 is the older Google video model, and it is being phased out. The model in the Nov 5, 2026 announcement, Cloud docs, and model card is Gemini Omni 1.2 Flash.
Use neither description as a lab result. “Near-flagship” and “better reference following” are the makers’ words.
Edited videos, if you are not calling an API
These two rows are generation models. A presenter video or a script-to-stock edit is a different product. Synthesia is the catalog pick for avatar explainers. Fliki is the catalog pick for a script with voiceover and stock visuals. Their plan prices were not re-checked for this comparison. The wider tool list is best AI video generators.
Sources
Read on Nov 17, 2026, unless a page’s own date is noted.
- Introducing Gemini Omni 1.2 Flash, Nov 5, 2026: https://blog.google/innovation-and-ai/technology/developers-tools/introducing-gemini-omni-1-2-flash/
- Gemini Omni 1.2 Flash on Google Cloud, model id, durations, resolutions, and C2PA: https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/gemini/omni-1-2-flash
- Gemini Omni 1.2 Flash generally available, Nov 5, 2026: https://cloud.google.com/blog/products/ai-machine-learning/gemini-omni-1-2-flash-generally-available/
- DeepMind model card, published Nov 5, 2026: https://deepmind.google/models/model-cards/gemini-omni-1-2-flash/
- Dreamina Seedance 2.5 series doc: https://docs.byteplus.com/docs/en/modelark/seedance-2-5.md
- BytePlus ModelArk pricing, updated 2026-11-16: https://ai.byteplus.com/ark/region:ap-southeast-1/docs/model-pricing
Pricing last checked Nov 17, 2026. The 4.4 marks are an editorial fit for the documented job. This page contains affiliate links. A commission does not change which model is recommended for which job.
Frequently asked questions
Which is better, Seedance 2.5 Flash or Gemini Omni 1.2 Flash?
There is no single winner. Seedance 2.5 Flash is the documented choice for a clip longer than 15 seconds, up to 30, and for a public reference set of 10 images, 5 videos, and 5 audios. Gemini Omni 1.2 Flash, model id gemini-omni-1.2-flash, generally available Nov 5, 2026, is the documented choice for 1080p or 4K and for Google's listed $0.07 per second at 720p. A single Omni generation is 3 to 15 seconds. Veo 3.1 is the older Google video model and is being phased out. This page is official docs and pricing pages, read Nov 17, 2026.
Is Seedance cheaper than Gemini Omni?
At 720p, the published per-second figures favor Gemini Omni 1.2 Flash. Google's Nov 5, 2026 posts list $0.07 per second for 720p. BytePlus's worked example for Seedance 2.5 Flash, 16:9, 5 seconds, no video input, is $0.76, or $0.15 per second. The meters differ. Seedance is billed in tokens, at $7.0 per million for 480p and 720p, and a video-input example costs more. Google's 30 percent cut is against Gemini Omni 1.1, which Google also calls Omni 1.1 Flash. That line is a vendor comparison with Google's previous Omni model.
How long a clip can each model make?
Seedance 2.5 Flash lists 4 to 30 seconds in one request. Gemini Omni 1.2 Flash lists 3 to 15 seconds for generation and editing, and 3 to 10 seconds when the shot uses a first and last frame. Google's announcement calls out native 15-second video, up from 10 seconds on Omni 1.1, in one-second steps. Extension on Omni adds 3 to 10 seconds onto up to the last 30 seconds of the input. A single 20 or 30 second generation is the Seedance cell.
How many reference images does Gemini Omni 1.2 Flash accept compared with Seedance?
The Cloud model page lists up to 10 images per prompt and up to 3 videos per prompt. Seedance 2.5 Flash lists 10 images, 5 videos, and 5 audios. Google's announcement also describes short motion clips of up to three seconds, plus audio cues. The same Cloud page marks the audio modality as not supported, and the Cloud blog says audio references are gated behind allowlisting. This page does not turn those three audio lines into one file cap.
Is Veo 3.1 still Google's current video model?
Veo 3.1 is the older Google video model, and it is being phased out. The current flagship video model is Gemini Omni 1.2 Flash, model id gemini-omni-1.2-flash, generally available Nov 5, 2026. The announcement, the Cloud model page, and the DeepMind model card are the pages this comparison uses for that model.
Did you test these models, and which one looks more realistic?
No. We did not generate video with either model. BytePlus describes Flash as near flagship Seedance 2.5 quality at about 2x the flagship speed. Google describes better reference following, more consistent shots, and more natural speech. Those sentences are vendor claims. The 4.4 marks mean each model has a documented job on this page.