October 28, 2026 · AI Radar
Let's set up a scene, entirely imaginary: a small e-commerce site selling towels in Bursa, Turkey, one o'clock at night. The owner is trying to get the words "Şimdi %30 İNDİRİMLİ" ("Now 30% OFF") onto a campaign image. The image tool adds the text, but the "Ş" has turned into an "S," and the dot on the "İ" is missing. After the third try, the owner gives up and adds the text by hand in Canva.
Here's why I set up this scene: the first promise of Nano Banana 2.1, released on October 6, is exactly this pain point. Google is introducing the model with "improved text rendering in images." The question I actually want answered isn't in the announcement: does it write Turkish characters correctly?
According to Google's Gemini API changelog, gemini-nano-banana-2.1 was released as a stable version on October 6 [A]. Google describes it as an update to Nano Banana 2 (technical name Gemini 3.1 Flash Image): in the same speed and cost class, with better image quality, results more faithful to the prompt, characters that stay the same across several rounds of editing, and better text [A].
These are the changes I think will be useful; they don't all carry the same weight.
Text. The model page says text rendering and infographic layout have improved [A]. In how many languages, and how accurately? It doesn't say. The image generation guide doesn't give a list of languages either; the examples include a Spanish translation, and Turkish doesn't appear [A]. An independent comparison had found that Nano Banana Pro, from the previous generation, sometimes dropped diacritical marks or put them in the wrong place [C]. That test was for the older version; where things stand in 2.1 isn't clear from the official sources.
Character consistency. You can provide up to 14 reference images in a single request; according to the model page, up to four of them can be characters and up to ten can be objects [A]. For anyone producing five different campaign images with the same mascot or the same model, I think this is the real news, more important than the text, even.
Wide aspect ratios. The 1:4, 4:1, 1:8 and 8:1 ratios were added; output is 1K, 2K and 4K [A]. 8:1 is handy for a website banner strip or a storefront banner. It also says the tiling artifacts seen at 2K and 4K in wide ratios in the previous version have been fixed [A]. One note: the small 512-pixel output isn't available in 2.1 [A].
What happens to the old model? gemini-3.1-flash-image went onto the deprecation list the same day, but Google's deprecation page has no shutdown date: "No shutdown date announced" [A]. An October 29 date is circulating on some sites; you won't find any such date on the official page, so there's no need to panic this week.
Source: Google, Gemini API pricing page, as of October 28, 2026.
That's everyone's first question, so I gave it its own section. The answer is different in the app and in the API, and there's some uncertainty on the app side too.
In the Gemini app, free accounts can generate images and edit photos, with a daily quota; the size of the quota isn't stated. The press reported that 2.1 also came to the Gemini app [C], but Gemini's own release notes have no separate entry for 2.1, and exactly which version runs behind free accounts isn't clear from the official sources. If you're curious about the model change on the free plan, see the post where we covered the October 9 switch to a single model.
On the Google AI Studio side, the pricing page says AI Studio usage is free in all regions where it's offered [A]; Turkey is on the list of supported regions [A]. The same page shows 2.1's free API tier as "Not available" [A]. So if you're going to write code and call it from the API, you'll need to turn on billing.
Here's the before-and-after comparison:
| Nano Banana 2 (3.1 Flash Image) | Nano Banana 2.1 | |
|---|---|---|
| 1K image (standard) | $0.067 | $0.0336 |
| 2K image | $0.101 | $0.0504 |
| 4K image | $0.151 | $0.113 |
| Smallest output | 0.5K (512 px) | 1K |
| 1:4, 4:1, 1:8, 8:1 ratios | No | Added |
| Status | Being deprecated, no shutdown date | Stable version |
Prices from the Gemini API pricing page, as of October 28, 2026 [A]. Some third-party lists show a lower figure for 4K; the figure on the official page is $0.113.
Let me be honest: there's no independent test comparing quality; all we have are Google's own claims. The tangible difference is in price and access.
Midjourney has no free plan, Basic starts at $10 a month, and its interface is in English and Korean. Its aesthetics are strong, but it has no official API, so you can't connect it to a store system. ChatGPT offers image generation on the free plan too, with limits. What sets Nano Banana 2.1 apart is that you can find the same model both in a free app and in an API at a little over 3 cents per image.
For a small business, I don't think the right question is "Which one is prettier?" It's "Which one places my product, with my text, without breaking it?" And you can only see that with your own product photo.
If cleaning up the background is your only concern, single-purpose tools like Photoroom still bring fewer surprises. And if you'd rather add the text by hand, Canva is right where it's always been; my contrarian view is that for most campaign images, having the model write the text is still an unnecessary risk.
According to Google's guide, every image generated with this model carries a SynthID watermark [A]. It's invisible to the eye, but Google's verification tool can detect that the image was generated with AI. Keep this in mind if you're thinking of presenting a product photo as a "real shot."
The Gemini API terms say Google won't claim ownership of the generated content; they add that it may generate the same or similar content for others and that you're responsible for how it's used [A]. On free services, the content you send may be used for product development and may be read by human reviewers; the terms say not to send sensitive or personal information to these services [A]. Think twice before uploading a photo with a customer's face. This isn't legal advice, just a reading of the terms.
The one thing you can try today: a Turkish text test
0/5A five-minute test; you'll see the result with your own product.
| Tool | What changed | Label | Who it helps |
|---|---|---|---|
| Google AI Studio | Nano Banana 2.1 stable version (October 6); 1K image $0.0336 | New model, Price/limits | Anyone connecting image generation to their own system |
| Google AI Studio | gemini-3.1-flash-image being deprecated, no shutdown date announced (October 6) | Shutdown | Anyone using the old model in their code |
| Gemini | Gemini 4 Argon generally available (October 13); details in this post | New model | Anyone handing long, complex tasks to Gemini |
| ComfyUI | v0.42.0-0.42.1 (October 18-19): Krea 3 Image node, FLUX 3 tags | Update | Anyone generating images on their own computer |
Google says text rendering has improved but doesn't list the supported languages; accuracy with Turkish characters isn't clear from the official sources. In every image you're going to publish, check the letters ç, ş, ğ, ı and İ one by one.
In the Gemini app, free accounts can generate images with a daily quota. In the API, this model has no free tier; as of October 28, 2026, a 1K image costs $0.0336 and a 4K image costs $0.113.
Under the Gemini API terms, Google doesn't claim ownership of generated content, and you're responsible for how it's used. The image carries a SynthID watermark. Be especially careful with real people's faces and other brands' products.
If you need support with product shoots or short videos, see the production services page; if you want to connect image generation to an automation, see the AI automation page. To see where you stand first: START YOUR FREE ANALYSIS .
AI Guide
GeminiChat, writing, image generation and sourced research reports with Deep Research; use inside Gmail and Docs on paid plans.
Google AI StudioGoogle's developer workspace where you try Gemini models in the browser, draft prompts and apps, and get a free Gemini API key.
MidjourneyGenerates highly aesthetic images and short videos from text and reference images; widely used for architectural concepts and for atmosphere, material and lighting studies.
ChatGPTAnswering questions, drafting emails and other text, summarizing documents, writing product descriptions and social media copy; limited deep research on the free plan.
PhotoroomRemoves the background from product photos, generates product scenes and backgrounds with AI, and prepares images in bulk for marketplace listings.
CanvaCreates social media graphics, posters and presentations from templates; generates images with AI, removes backgrounds, and makes videos and short video clips from images.
ComfyUIRuns open image and video models on your own computer with node-based workflows; used for product images, image editing and video generation.Beyond GPT-6, ChatGPT got five new features in October 2026. Audio uploads are on paid plans, Slides on Pro and business plans, locked chats for everyone; Finances is US-only.
+2OpenAI released GPT-6.1 Terra on October 26, 2026: on the API, one million input tokens cost $1 and output tokens $5, half the price of its predecessor. It's available in ChatGPT Work and Codex on Plus and higher plans.
+1Google has deprecated Gemini 2.5 Pro and Flash; starting December 1, Gemini API requests are quietly redirected to 3.8 Flash and 3.5 Flash-Lite. What happens to price and quality, and which automations are affected?
+2We use cookies to improve the site and measure advertising performance. Analytics and advertising cookies run only if you accept them. Details are in our Privacy Policy and Data Protection Notice.
