Creating Images and Videos with Gemini (Nano Banana, Gemini Omni)

November 4, 2026

Updated:

Guide
Creating Images and Videos with Gemini (Nano Banana, Gemini Omni)

What is Creating Images with Gemini?

Creating images with Gemini is typing a description in words (a prompt), possibly with reference images, so that the Gemini app or Gemini API generates new images or edits existing ones; creating videos works the same way. The general introduction is in the post What is Gemini?. The most important thing to say first: model names have changed, Imagen and Veo are now old names. According to official documentation read on 04/11/2026, Imagen has stopped working on Gemini API and is replaced by the Nano Banana family; the video section in the Gemini app now revolves around Gemini Omni, while Veo 3.1 is still introduced by Google DeepMind but Veo preview versions on Gemini API have reached their end date. This post does not promise images or videos ready to use immediately for advertising, and is not legal advice. The post covers current model names, limits by plan, SynthID watermark, content policy, copyright, prompt writing tips and real experience with many OPCAIOS errors still remaining.

Diagram of creating images and videos with Gemini: prompts and reference images go into the Gemini app or Gemini API; images made by the Nano Banana family, videos made by Gemini Omni or Veo; all products bear hidden SynthID watermark and C2PA metadata, visible watermark mandatory in Vietnam except AI Ultra plan; users check content policy and usage rights before posting
Creating images and videos with Gemini: models, watermark and steps to check before use. Diagram: OPCAIOS.

What are Imagen, Veo and Nano Banana called now?

Nano Banana is the trade name for the family of image generation and editing models built on Gemini, while Gemini Image is the technical name for the same family; Imagen is the older line of image models that has stopped on Gemini API. According to the Deprecations page of Gemini API (read on 04/11/2026), imagen-4.0-generate-001, Ultra and Fast versions have the earliest stop date of 17/08/2026, the replacement model listed is gemini-nano-banana-2.1. The Imagen page in the documentation states directly that Imagen has stopped, use Nano Banana instead. The table of image and video model names readable today:

Trade name Model code on API Notes from documentation
Nano Banana 2.1 gemini-nano-banana-2.1 Newest, released 06/10/2026, better text in images, maintains consistency through multiple edits
Nano Banana 2 gemini-3.1-flash-image Previous version, documentation recommends switching to 2.1
Nano Banana 2 Lite gemini-3.1-flash-lite-image Cheapest and fastest, not optimized for many reference images
Nano Banana Pro gemini-3-pro-image 4K images, complex layouts, accurate text
Nano Banana gemini-2.5-flash-image Old version, earliest stop date 15/03/2027
Gemini Omni Flash gemini-omni-1.1-flash Generate, edit, extend videos, with audio
Imagen 4 imagen-4.0-generate Discontinued

The Models page of Gemini API read on 04/11/2026 no longer lists Veo; the only video model there is Gemini Omni Flash. Many people still search by the names Imagen and Veo, but newcomers should learn straight to the Nano Banana family for images and Gemini Omni for videos.

Can Veo still be used, what is Gemini Omni?

Gemini Omni is a tool for creating and editing videos through conversation in the Gemini app, launched on 19/05/2026 according to the release notes page. The release applies to people 18 and older with a Google AI plan, scope varies by region. As for Veo, three official sources read on 04/11/2026 do not say exactly the same thing:

  1. The Veo page of Google DeepMind still calls Veo 3.1 the leading video model, with native audio, outputs 1080p and 4K, sample videos 8 seconds long, has a try button in Gemini, Flow and Gemini API.
  2. The plans page of the Gemini app lists Veo 3.1 Lite as a limited trial form for the Google AI Pro plan, alongside Gemini Omni Flash.
  3. The Deprecations page lists the preview versions veo-3.1-generate-preview, Fast and Lite versions have the earliest stop date of 22/10/2026, replaced by gemini-omni-1.1-flash; official Veo versions remain on Gemini Enterprise Agent Platform.

Practical conclusion: with the Gemini app, look for Omni; for programmers, check the models page on the day of deployment. For API details see the post Gemini API and AI Studio.

How many images and videos does each Gemini plan create?

The creation limits for images and videos in the Gemini app are no longer fixed numbers but calculated as multiples of a standard level, with video being the most quota-consuming type. According to the "Gemini Apps limits & feature availability" page (accessed 04/11/2026), with no plan you use the standard level, AI Plus is 2x, AI Pro is 4x, and AI Ultra is 5x or 20x AI Pro depending on the plan; the limit resets every 5 hours until it hits the weekly limit and can change without notice. The gemini.google plan page shows for Vietnam region on 04/11/2026:

Plan Price shown (VND/month) Images and video per page
Free 0 Create images with Nano Banana 2; no video creation
Google AI Plus 132,000 2x, video creation, 200 Google Flow credits
Google AI Pro 489,000 4x, Nano Banana Pro, 1,000 Flow credits, Veo 3.1 Lite trial
Google AI Ultra From 2,250,000 Highest level, 10,000 or 25,000 Flow credits

Prices can change and may differ when purchased through app stores. More detailed analysis of each plan in the article Gemini free and paid.

How do SynthID watermark and visible watermark work?

SynthID is a hidden watermark developed by Google DeepMind, embedded in images, videos and audio created by Google AI, invisible to the human eye but recognizable by Google tools. According to the help page "Manage watermark settings in Gemini Apps" (accessed 04/11/2026), all content created or edited using Gemini apps has SynthID and Content Credentials metadata (C2PA). Gemini API image documentation also states that all generated images have SynthID. Visible watermark is a separate matter, with four points:

  1. On web, go to Settings, Media Watermark to turn on or off; this setting only controls the visible watermark.
  2. Turning off the visible watermark does not remove SynthID and C2PA.
  3. In India, South Korea and Vietnam, the option to turn off is only available for AI Ultra plan; other users in these countries always have the visible watermark.

What does content policy block?

The content policy of the Gemini app is a set of general principles applied to text, images and video, with no separate rules for image creation. The policy guidelines page of gemini.google (accessed 04/11/2026) lists six categories:

  1. Content threatening child safety.
  2. Dangerous activities, such as instructions for self-harm or weapon making.
  3. Violence and gore that is shocking or gratuitous.
  4. Misinformation that could harm health, safety, finance.
  5. Harassment, incitement, discrimination.
  6. Sexual content.

Additionally, Google's Terms of Service prohibit using AI content to make others believe it was created by humans.

Can I use Gemini images for advertising?

This section is for reference only, not legal advice. Google's Terms of Service (effective 30/07/2026, accessed 04/11/2026) state that some services allow creation of original content and Google does not claim ownership of that content. Not claiming ownership does not mean the image is necessarily protected by copyright, nor does it guarantee the image does not resemble someone else's work. The Gemini API terms also state Google may create content that is the same or similar for others. Five steps to verify before using an image in advertising:

  1. The reference image you upload must be an image you have the right to use.
  2. No logos, characters, or faces of third parties.
  3. Do not misrepresent the product's function.
  4. Keep the watermark and metadata, clearly note the image was AI-created when the advertising platform requires it.
  5. Read the ad policy of the platform you plan to run on; we have not verified Vietnam regulations on labeling AI content in advertising.

How a marketing person works with AI, see the article one-person marketing agency with AI.

How do I write a prompt to generate images with Gemini so they actually turn out usable?

A good image generation prompt is a specific scene description, not a string of scattered keywords. Gemini API's image generation documentation (read 04/11/2026) recommends describing the scene in real detail; for realistic images, mention frame type, subject, background, lighting, camera angle and lens; for illustrations, mention style, material, line work, color palette; then refine gradually through multiple conversation turns. Five tips to apply:

  1. Write in order: subject, action, background, lighting, camera angle, style.
  2. State aspect ratio for where you're posting, for example 9:16 for vertical video, 1:1 for square posts.
  3. If you want text in the image, write it out clearly with position, then check your Vietnamese spelling with diacritics.
  4. Keep characters consistent using reference images; the documentation says max 14 images, Nano Banana 2.1 accepts up to 4 character images.
  5. Adjust one thing at a time each turn: "keep everything the same, just change the background to cream color".

Example for a bakery: "Close-up shot of a baguette fresh from the oven on a wooden table, morning sunlight from the left window, 45 degree camera angle, 50mm lens, blurred background, 4:5 ratio."

OPCAIOS's experience

OPCAIOS has an internal short video production tool. During a test run on 03/11/2026, this tool used a Gemini Flash Image generation model to create illustrations for each scene. The results had errors: some images repeated between scenes, and one image drew a character wrong. Three takeaways:

  1. Batch image generation from a script requires a review step before each image goes into the video.
  2. Keeping characters consistent across multiple scenes is the hardest part; you need fixed reference images instead of just text description.
  3. Model names in source code need to be checked against the Deprecations page, because the Gemini family changes model codes pretty fast.

We haven't used Veo or Gemini Omni for real video yet, we haven't measured cost per image, and we don't have ad effectiveness data from AI images.

Frequently asked questions

Is image generation with Gemini free?

It is at the basic level. According to the app limits page for Gemini read on 04/11/2026, image generation with Nano Banana 2 works even without a paid plan, but the rate limit is standard, resets every 5 hours and has a weekly cap. Re-generating images with Nano Banana Pro and video generation need Google AI Plus tier or higher.

Does Imagen still work?

Not on Gemini API: the Imagen page lists Imagen models as discontinued and recommends using Nano Banana, the Deprecations page shows Imagen 4's earliest shutdown date as 17/08/2026. Google DeepMind's pricing page and models page read on 04/11/2026 no longer mention Imagen.

Can I remove the watermark from Gemini images in Vietnam?

According to the watermark help page read on 04/11/2026, in Vietnam only AI Ultra tier users see the option to turn off watermark display. Even if turned off, hidden SynthID and C2PA metadata remain, and Google has tools to detect them. Using outside tools to remove the mark doesn't make the image stop being an AI image, and it may violate terms about not deceiving viewers.

Google's terms say Google doesn't claim ownership of content you create, but that doesn't answer whether the image gets legal protection, because it depends on each country's law. Google can also generate similar images for other people. This is not legal advice; for major ad campaigns or brand marks, ask an intellectual property lawyer.

Where should a newcomer start generating images with Gemini?

Start in the Gemini app with the plan you have, pick one concrete task like a post cover image or product illustration. Write your prompt in order: subject, background, lighting, camera angle, style, then refine one thing at a time through multiple turns. If you want to systematically learn how to put AI images and video into your content workflow, check out OPCAIOS's AI application training course.

Source: Gemini API docs (models, image-generation, imagen, deprecations), Google DeepMind (models page, Gemini Image, Veo), gemini.google (subscriptions, release notes, policy guidelines), Gemini Apps Help (limits & feature availability, Manage watermark settings), Google Terms of Service (read 04/11/2026). Not verified: exact number of images and videos per plan, max video length for Gemini Omni in app, Vietnam regulations on AI labeling in ads, image pricing on Gemini API. Experience is OPCAIOS's own.

Translated from Vietnamese with AI. Read the original.

Want to apply this in your business?

Book a free 30-minute call (Vietnam time, in English or Vietnamese).

Book a call