I'd been subscribing to Gemini Pro without ever thinking about creating AI images separately. When you generate images in the Gemini app, a watermark gets slapped right on them, and removing it means paying for an even more expensive plan. So I'd been just making do with Canva, until one day the YouTube algorithm recommended something called "Google Flow."

At first, I clicked in with zero expectations, thinking "what's this?" But after messing around with it for about 40 minutes, I was genuinely surprised. I could generate Nano Banana Pro quality images without watermarks using just the Gemini Pro subscription I was already paying for. No additional charges.

But looking at articles floating around online, a lot of them have titles like "Unlimited free!" In reality, after using it myself, it's not unlimited. I'll go into detail about this later.

Current AI Image Generation Options — All Somewhat Lacking

Here's an honest rundown of the options I've tried:

Gemini App. Image generation works, but the free tier gives you 2 images per day at 1K resolution with a visible watermark. Subscribing to Google AI Pro ($19.99/month) removes the watermark and bumps resolution up to 2K, but the daily usage limit remains.

Google AI Studio. It's a developer platform, so it's not very user-friendly for regular users, and the free tier only gives you basic Nano Banana (Gemini 2.5 Flash Image). Nano Banana Pro (Gemini 3 Pro Image) requires a paid API key, and at about $0.15 per 4K image, costs add up quickly if you use it a lot.

ChatGPT Image Generation. Quality is good. But the free tier gets throttled with heavy use, and even paid plans can slow down depending on usage.

Everything felt like "this is decent enough, but... there's always one thing missing."

How I Discovered Google Flow

Google Flow is an AI creative studio operated by Google Labs. Originally it was a Veo (video generation)-centric platform, but recently Nano Banana image generation was integrated as a core feature, allowing you to work on both images and videos in one place.

The URL is labs.google/flow or flow.google. Some articles online list labs.google.com/fx/tools/flow, but the URL seems to have changed, so just searching for "Google Flow" is the most reliable approach.

What's Actually Good About It

Nano Banana Pro quality at no additional cost. Google AI Pro subscribers ($19.99/month) can generate images directly with the Nano Banana Pro model on Flow. Free users can only use the basic Nano Banana.

No visible watermark. Pro and above subscribers don't get a visible watermark. However, an invisible SynthID digital watermark is embedded in all images. This is metadata-level tracking that Google uses to trace AI-generated images — it's invisible to the naked eye and doesn't affect image quality. Honestly, this was what I liked most.

2 variations per prompt. Enter one prompt and you get two versions to compare. There are subtle differences, so you can immediately pick "ah, this one's better."

Reference image upload. You can upload logos, brand assets, or reference photos alongside your prompt. This helps reduce the problem of AI interpreting your logo however it wants and drawing something weird. This was genuinely useful for making YouTube thumbnails and blog banners.

Download resolution. 1K (free/basic), 2K (Pro), 4K (Ultra). Even 1K was plenty for thumbnails.

By the way, Google recently launched Nano Banana 2 (Gemini 3.1 Flash Image), which combines the quality of Nano Banana Pro with Flash's speed. Flow has started supporting Nano Banana 2 as well, and apparently character consistency and text rendering have improved significantly. I haven't used Nano Banana 2 much yet, so I'll write a detailed comparison separately later.

How to Actually Use It — Not That Hard

  1. Log in at labs.google/flow with a Google account linked to your Gemini paid subscription.
  2. Create a new project, and Nano Banana Pro should be set as the default model.
  3. Write your prompt. The more specific you are about style, color, composition, text, etc., the better the results.
  4. Attach reference images if needed.
  5. Hit enter and 2 variations are generated. Download the one you like at your preferred resolution.

I tried it for making YouTube thumbnails, and the images were much more unique than what I'd get picking templates in Canva. The text rendering in Nano Banana Pro was noticeably good, especially for thumbnails with text. Other AI image generators often mess up even English text, but here it came out almost perfectly. Only for English though.

"Unlimited Free" Is Not Accurate

Let me be honest about something here.

There are quite a few articles online saying "Generate AI images for free unlimited on Google Flow!" — this is not accurate.

Here are the facts:

What Google AI Pro ($19.99/month) subscribers get on Flow is 1,000 AI credits per month. How many credits each image consumes varies by resolution, but "unlimited" is definitely not the case. Heavy use will drain your credits.

Going to the Ultra tier ($249.99/3 months, effectively ~$125/month) bumps it up to 25,000 credits, but that's not unlimited either.

I've seen reviews saying "I made hundreds of images without hitting any limit," but they were probably within the Pro credit range or the policy was different in the early days. Google changes their credit policies really frequently. In November 2025, the free tier daily limit was reduced from 3 images to 2.

So to be precise, this is "utilizing a feature included in the subscription you're already paying for." Rather than "free," "at no additional cost" is the accurate expression. It might seem like a minor distinction, but not knowing this can lead to "wait, I thought it was unlimited?" disappointment.

Cost Comparison — Honestly Not Bad

Despite the limitations, the value proposition is decent.

Using Nano Banana Pro directly via API on AI Studio costs about $0.15 per 4K image. 100 images = $15. But if you're already on a Pro subscription ($19.99/month), you can create a significant volume on Flow at no extra cost, so it's clearly a win.

I was also paying for Canva Pro ($12.99/month), and after discovering Flow, honestly, the reason to keep my Canva subscription became a bit questionable. Of course, Canva's strength is in templates and editing features so it's not completely replaceable, but for the purpose of "creating the image source itself," Flow is overwhelmingly better. I played around for 40 minutes and cranked out 6 YouTube thumbnails. The same task on Canva would have taken at least 2 hours.

Limitations and Shortcomings

UI isn't image-dedicated. Since Flow was originally a video generation platform, it can feel a bit heavy when you just want to quickly churn out images. If you only want image generation, the Gemini app might actually be lighter and more intuitive.

Korean text. This is a bit of a dealbreaker — while English text rendering is quite accurate, results are still unstable when trying to create images with Korean text. Characters sometimes break or get mixed with nonexistent Chinese character-like shapes. For Korean thumbnails, it's safer to add the text separately afterward.

Confusing subscription names. Google renamed the subscription tiers in 2025, so "Gemini Advanced" became "Google AI Pro," and a cheaper tier called "Google AI Plus" (~$7.99/month) was added. Many articles online still say "Gemini Plus," but the current official name is "Google AI Pro." The price remains the same at $19.99/month.

Concurrent prompt throttling. Being able to run multiple prompts simultaneously is an advantage, but during peak hours, generation speed noticeably slows down. In Korean time, early morning is fastest, and afternoons take longer.