# Gemini Image Generation Guide: Pick the Right Tool and Model

> As of October 8, 2026, the free Gemini app runs Nano Banana 2, Google AI plans get Nano Banana 2.1, and API images are billed: $0.0336 per 1K image on 2.1.

- URL: https://blog.laozhang.ai/en/posts/gemini-image-generation-complete-guide
- Published: 2026-03-26
- Updated: 2026-10-08
- Author: LaoZhang AI Team (https://blog.laozhang.ai/en/about)
- Topic: AI Image Generation
- Tags: Gemini, Gemini Image Generation, Nano Banana 2.1, Nano Banana 2, Nano Banana Pro, Gemini API, Interactions API

---
Gemini makes images with a family of models Google calls Nano Banana. For most people, the quickest route is the Gemini app: go to [gemini.google.com](https://gemini.google.com), click **Open Sidebar**, then **Images**, and describe what you want. Without a paid plan, that runs Nano Banana 2 and downloads at 1K. With a Google AI plan (Plus, Pro, or Ultra), you get Nano Banana 2.1, Google's newest image model, and 2K downloads ([Gemini image page](https://gemini.google/overview/image-generation/), [Gemini Apps Help](https://support.google.com/gemini/answer/14286560?hl=en)). Google doesn't publish an image count for either. Your allowance refreshes every 5 hours until you reach a weekly cap.

Developers reach the same models through the Gemini API. Google recommends `gemini-nano-banana-2.1` for new projects, the recommended interface is the Interactions API, and image models have no free tier. As of October 8, 2026, a Nano Banana 2.1 image costs $0.0336 at 1K, $0.0504 at 2K, and $0.113 at 4K. Imagen, Google's older text-to-image line, is shut down on the Gemini API, and calls to it no longer work.

What changed on October 6, 2026: Nano Banana 2.1 became generally available, and on the Gemini API, Nano Banana 2 (`gemini-3.1-flash-image`) is now deprecated with no shutdown date announced ([Gemini API changelog](https://ai.google.dev/gemini-api/docs/changelog)). Existing Nano Banana 2 code keeps running. The API examples below now use 2.1, and the [migration notes](#moving-code-from-gemini-3-1-flash-image-to-gemini-nano-banana-2-1) cover what else to change.

## Which Google tool should you use for images?

Google offers image generation in five places. They differ in which model you get, how the limit is counted, who pays, and who is allowed in. Limits and plan details below are as of September 26, 2026; Gemini app and API models are as of October 8, 2026.

| Where | Model you get | Without a paid plan | With a paid plan | Who can use it |
| --- | --- | --- | --- | --- |
| Gemini app (web and mobile) | Nano Banana 2 without a plan; Nano Banana 2.1 with a Google AI plan; Nano Banana 2 Lite on Flash-Lite | "Standard limits": compute-based, refresh every 5 hours up to a weekly cap, no image count | Nano Banana 2.1; AI Plus 2x, AI Pro 4x, AI Ultra 5x or 20x AI Pro's limits; Redo with Nano Banana Pro | 13+ to generate, 18+ to edit (personal accounts), in supported countries |
| AI Mode in Google Search | Google doesn't name the default model | Up to 25 images per 24 hours | Up to 50 (Plus), 100 (Pro), 1,000 (Ultra) | 18+, in listed countries |
| Google Flow | Nano Banana 2 Lite at no charge; Nano Banana 2 and Pro spend credits | 50 credits a day, no rollover | Extra monthly credits: +200 (Plus), +1,000 (Pro), +10,000 or +25,000 (Ultra) | 18+, in listed countries |
| Google AI Studio (web) | Nano Banana is listed among the models an AI Pro plan unlocks | "Modest quota", with no free image count published | Higher quota, inside the AI Studio web interface only | Google account |
| Gemini API | Whichever of five model IDs you call; Google recommends Nano Banana 2.1 | No free image tier | Pay per image | Project with billing set up |

Sources: [Gemini Apps limits](https://support.google.com/gemini/answer/16275805?hl=en), [Gemini image page](https://gemini.google/overview/image-generation/), [AI Mode image creation](https://support.google.com/websearch/answer/16649374?hl=en), [Flow credits](https://support.google.com/flow/answer/16526234?hl=en), [AI Studio plans](https://ai.google.dev/gemini-api/docs/google-ai-plans), [Gemini API pricing](https://ai.google.dev/gemini-api/docs/pricing).

The choice usually comes down to one question each:

- **You just want pictures, edits of your own photos, or quick memes:** the Gemini app. It edits photos you upload, and you can keep refining the result in the same chat.
- **You want to know exactly how many free images you have left:** AI Mode in Google Search (AI Mode, then **+**, then **Create images**). It is the only Google consumer product that prints a per-day count. In the US, English-language users 18 and older can also pick Nano Banana Pro there, with its own daily limit.
- **You are building a project that mixes images and video:** Google Flow. Its no-charge image model is Nano Banana 2 Lite. Google's help pages disagree on which model Flow selects by default, so check the model and its credit cost in the prompt box settings before you generate.
- **You want to test prompts before writing code:** Google AI Studio. Google lists Nano Banana among the models an AI Pro plan unlocks there, so a free account shouldn't count on free image generation in AI Studio. AI plan benefits also stop at the AI Studio web interface; API calls are billed separately.
- **You need images inside an app, a script, or a batch job, or you need 4K:** the Gemini API.

In the US, Google AI Plus costs $4.99 a month, AI Pro $19.99, and AI Ultra $99.99 or $199.99 depending on the tier ([Google subscriptions page](https://gemini.google/us/subscriptions/?hl=en)). Any of the three unlocks Nano Banana 2.1 in the Gemini app. None of them pays for API usage.

## Nano Banana names and the model IDs behind them

The Gemini app shows brand names, while code needs model IDs, and the app never displays an ID. That is why the app says "Nano Banana 2" while your code says `gemini-3.1-flash-image`: they're the same model. The newest one breaks the old naming pattern. Nano Banana 2.1 has no "Gemini 3.x Flash" name, and its ID is simply `gemini-nano-banana-2.1`.

| Name in the app | Model ID in the API | How to get it in the Gemini app | API price per image | Output sizes |
| --- | --- | --- | --- | --- |
| Nano Banana 2.1 | `gemini-nano-banana-2.1` | Any Google AI plan (Plus, Pro, Ultra) | $0.0336 (1K), $0.0504 (2K), $0.113 (4K) | 1K, 2K, 4K |
| Nano Banana 2 Lite | `gemini-3.1-flash-lite-image` | Set Gemini to Flash-Lite | $0.0336 (1K) | 1K only |
| Nano Banana 2 (deprecated in the API) | `gemini-3.1-flash-image` | Default without a plan (Flash or Pro in the model picker) | $0.045 (0.5K), $0.067 (1K), $0.101 (2K), $0.151 (4K) | 512px, 1K, 2K, 4K |
| Nano Banana Pro | `gemini-3-pro-image` | Paid plan: **Redo with Pro** on an image you made | $0.134 (1K or 2K), $0.24 (4K) | 1K, 2K, 4K |
| Nano Banana (legacy) | `gemini-2.5-flash-image` | Not offered | $0.039 | About 1024px |

Prices are Google's standard paid rates per output image as of October 8, 2026 ([pricing page](https://ai.google.dev/gemini-api/docs/pricing)); input text, reference images, and thinking are billed as tokens on top.

What each one is for, in Google's own framing ([image generation docs](https://ai.google.dev/gemini-api/docs/image-generation), [Nano Banana 2.1 model page](https://ai.google.dev/gemini-api/docs/models/gemini-nano-banana-2.1)):

- **Nano Banana 2.1** is an update to Nano Banana 2 that keeps Flash-level speed and cost, with better visual quality, prompt adherence, text rendering, and character consistency across edits. It takes up to 14 reference images (up to 10 objects and up to 4 characters), supports grounding with Google Web and Image Search, and lets you set thinking to minimal, medium (the default), or high. Google recommends it for all new projects.
- **Nano Banana 2 Lite** is the fastest and cheapest. Google says it is not optimized for multiple reference images or for editing over several turns, and it doesn't support grounding with Google Search.
- **Nano Banana 2** was Google's go-to image model until October 6, 2026. It is now deprecated on the Gemini API with no shutdown date, and it is still the only current model that outputs 512px images. On Vertex AI, Google lists support until May 28, 2027 or later.
- **Nano Banana Pro** is for the hardest jobs: dense text, infographics, translations inside an image, brand consistency. In the Gemini app it's worth a redo when an image gets the text or small details wrong.
- **Nano Banana (2.5)** is deprecated, and Google's pages disagree on when it shuts down: the [deprecations page](https://ai.google.dev/gemini-api/docs/deprecations) says March 15, 2027, while the pricing page still warns of October 2, 2026. Google's listed replacement is Nano Banana 2 Lite.

### Which Gemini image model should you use?

In the Gemini app the plan decides: free accounts get Nano Banana 2, Google AI plans get 2.1, and **Redo with Pro** is the manual step up. In the API, you pick:

- **Start new projects on Nano Banana 2.1.** It costs the same as Lite at 1K ($0.0336) and less than Nano Banana 2 at every size it offers.
- **Use Nano Banana 2 Lite for high-volume 1K drafts.** The per-image price matches 2.1 at 1K, but its tokens are cheaper: $0.25 per million input tokens and $1.50 per million thinking tokens, against $1.50 and $7.50 on 2.1.
- **Use Nano Banana Pro when exact wording matters.** In a [9-prompt test of Nano Banana 2.1, Pro, and GPT Image 2.5](https://blog.laozhang.ai/en/posts/nano-banana-2-1-vs-pro-vs-gpt-image-2-5) run at 1K with one draw per prompt on October 8, 2026, Pro kept the requested wording while 2.1 changed it in 3 of the text tasks. Pro costs $0.134 per image at 1K or 2K.
- **Stay on Nano Banana 2 only for 512px output** or a pipeline you can't change yet. Plan the move to 2.1 even without a deadline.

![Conceptual guide matching new projects, fast 1K drafts and text-focused images to Nano Banana 2.1, 2 Lite and Pro, with their API IDs and a 512px exception.](https://blog.laozhang.ai/posts/en/gemini-image-generation-complete-guide/img/choose-gemini-image-model.webp)

For per-call cost on 2.1 including thinking tokens, see [Nano Banana 2.1 API Pricing](https://blog.laozhang.ai/en/posts/nano-banana-2-1). If you're choosing between 2.1, Lite, Pro and the deprecated Nano Banana 2 for a specific job, the [Nano Banana 2 Lite vs 2 vs 2.1 vs Pro comparison](https://blog.laozhang.ai/en/posts/nano-banana-pro-vs-nano-banana-2) goes through it task by task.

## Make your first image in the Gemini app

Google documents these steps for the computer version of the app; the Android and iPhone apps have their own layout, so a label may sit elsewhere.

1. Sign in at [gemini.google.com](https://gemini.google.com) with your Google account.
2. Without a plan, keep **Flash** or **Pro** in the model picker for Nano Banana 2, or choose **Flash-Lite** when you want a fast rough draft. With a Google AI plan, the app uses Nano Banana 2.1.
3. Click **Open Sidebar**, then **Images**. Google's Gemini image page also points to **Create images** in the tools menu. Pick a template or type a prompt such as "Create an image of a golden retriever surfing at sunset, in a splashy watercolor style."
4. Refine in the same chat instead of starting over: "Keep everything the same, but make the sky stormy."
5. To download, hover over the image and click **Download full size**. Google's help page says you get a 1K file without a Google AI plan and 2K with one; its Gemini image page lists 2K for both Nano Banana 2 and 2.1, so check the pixel size of your file. **More**, then **Export to Docs** sends it to Google Docs.

To edit a photo you already have, upload it in the chat, or open **Library** in the sidebar, click an earlier image, click **Chat**, and type what to change. You can also upload several images and ask Gemini to combine them into a new one. Editing requires you to be 18 or older on a personal account.

On a paid plan, a **Redo with Pro** option appears under **More** (the three-dot menu) at the bottom right of an image. It regenerates the picture with Nano Banana Pro, which helps most with text and infographics. It has a catch: once you reach your daily image quota, you can't redo any more with Pro either. Google doesn't say how large that daily quota is.

Every image carries an invisible SynthID watermark. As of late August 2026, eligible accounts could turn off the visible watermark under Settings, then **Media watermark**; work and school accounts keep it. The steps and eligibility rules are in [how to turn off Gemini's visible watermark](https://blog.laozhang.ai/en/posts/gemini-image-generation-error-limit-watermark).

### How many images you can make in the app

There is no fixed number. Google published a figure of 100 free images a day in 2025, but the current help pages no longer print any image count. Instead, the app uses "compute-based usage limits" that factor in the complexity of your prompt, the models and features you use, and the length of your chat. A paid plan multiplies the standard limit (2x for Plus, 4x for Pro), and it doesn't turn into a number of images either. Google's Gemini image page says paid plans get "higher generation limits" with 2.1, again without a number.

To see where you stand, open [gemini.google.com/usage](https://gemini.google.com/usage) (Settings, then **Usage limits**). It shows **Current usage** as a percentage with the time it resets, and a separate **Weekly limit**. In one test on a free account in September 2026, a single simple image moved the current-usage meter by 2 to 3 percentage points and left the weekly meter at 0%. That is one sample, not a reliable count per window; complex prompts, long chats, and Pro-level models draw more.

When you hit the limit, you can wait for the 5-hour refresh (or the weekly reset if that's the one you've used up), switch to AI Mode for its fixed 25 images a day, or upgrade. The [Gemini free image limit guide](https://blog.laozhang.ai/en/posts/gemini-image-generation-free-limit-2026) compares every free allowance, including upload limits, and [Can You Use Nano Banana for Free?](https://blog.laozhang.ai/en/posts/is-nano-banana-free) lists the no-cost routes.

## Generate images with the Gemini API

The code below follows the request patterns in Google's [image generation docs](https://ai.google.dev/gemini-api/docs/image-generation), with the model set to `gemini-nano-banana-2.1`, which those docs recommend for all new projects. It uses the Interactions API, which became generally available in June 2026 and is what Google recommends for new projects. Code you already have that calls `generate_content` keeps working, and Google says that API remains fully supported, but it now counts as legacy, and new models and features launch on the Interactions API first ([Interactions overview](https://ai.google.dev/gemini-api/docs/interactions)).

### Set up billing and a key

Image models have no free tier in the API, so the project behind your key needs billing before the first image call. In [Google AI Studio](https://aistudio.google.com), copy the key from the API keys page (or click **Create API key**), then click **Set up billing**. Google's billing dialog asks for a minimum $5 prepayment in paid credits. Then:

```bash
pip install -U google-genai
export GEMINI_API_KEY="YOUR_API_KEY"
```

The SDK reads `GEMINI_API_KEY` from the environment, so the key never has to appear in your code.

### First call: set the aspect ratio and size

```python
import base64
from google import genai

client = genai.Client()  # reads GEMINI_API_KEY

interaction = client.interactions.create(
    model="gemini-nano-banana-2.1",
    input=(
        "Create a photorealistic product shot of a matte white ceramic mug "
        "on a walnut desk, soft morning window light, shallow depth of field"
    ),
    response_format={
        "type": "image",
        "mime_type": "image/jpeg",
        "aspect_ratio": "16:9",
        "image_size": "2K",
    },
)

image = interaction.output_image
if image is None:
    raise SystemExit("No image in the response. Rephrase the prompt and try again.")

with open("mug.jpg", "wb") as f:
    f.write(base64.b64decode(image.data))
```

What this request does:

- `interaction.output_image` returns the last image in the response as base64 data.
- A 16:9 image at 2K on Nano Banana 2.1 comes out at 2752 × 1536 pixels and costs $0.0504 at the standard rate, plus input and thinking tokens.
- Leave out `response_format` and you get a 1K image at 1:1 (or matching the size of an image you sent), plus some text from the model. With an image-only `response_format`, the text is dropped.
- Nano Banana 2.1 and Nano Banana 2 accept 14 aspect ratios: 1:1, 2:3, 3:2, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9, 21:9, plus the extra-tall and extra-wide 1:4, 4:1, 1:8, and 8:1. Google says 2.1 fixes the tiling artifacts those four wide and tall ratios showed at 2K and 4K. Nano Banana Pro and Nano Banana 2 Lite take the first ten.

The same request over REST:

```bash
curl -s -X POST \
  "https://generativelanguage.googleapis.com/v1beta/interactions" \
  -H "x-goog-api-key: $GEMINI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gemini-nano-banana-2.1",
    "input": "Create a photorealistic product shot of a matte white ceramic mug on a walnut desk",
    "response_format": {
      "type": "image",
      "aspect_ratio": "16:9",
      "image_size": "2K"
    }
  }'
```

### Edit in follow-up turns or from your own photo

Google recommends multi-turn conversation as the way to iterate on an image. Pass the previous interaction's ID and describe only the change:

```python
follow_up = client.interactions.create(
    model="gemini-nano-banana-2.1",
    input="Keep everything the same, but make the mug matte black.",
    previous_interaction_id=interaction.id,
    response_format={"type": "image", "aspect_ratio": "16:9", "image_size": "2K"},
)
```

This works because the Interactions API stores each request by default (`store=true`): for 55 days on the paid tier and 1 day on the free tier. If you set `store=False` for privacy, `previous_interaction_id` stops working, and you need to send the image back yourself.

![Conceptual Gemini Interactions API workflow from billing and explicit image settings to saving an image, then editing by stored interaction ID or by resending the image.](https://blog.laozhang.ai/posts/en/gemini-image-generation-complete-guide/img/gemini-request-save-edit.webp)

To edit a photo from disk, send a list that holds the instruction and the image:

```python
with open("room.png", "rb") as f:
    photo = base64.b64encode(f.read()).decode("utf-8")

edited = client.interactions.create(
    model="gemini-nano-banana-2.1",
    input=[
        {"type": "text", "text": "Replace the sofa with a green velvet one. Keep the lighting and camera angle."},
        {"type": "image", "data": photo, "mime_type": "image/png"},
    ],
)
```

Add more `image` items to use reference images, up to 14 in total. Only upload images you have the rights to use.

### Moving code from gemini-3.1-flash-image to gemini-nano-banana-2.1

There's no hard deadline: Google deprecated Nano Banana 2 on the Gemini API when 2.1 launched and hasn't announced a shutdown date ([deprecations page](https://ai.google.dev/gemini-api/docs/deprecations)). An October 29, 2026 date appeared on that page on launch day and was gone by October 7. The examples above differ from Nano Banana 2 code only in the model string. Four things can still catch you:

- **No 512px output.** Nano Banana 2.1 makes 1K, 2K, and 4K only. Send `"1K"` instead of `"0.5K"`, or keep 512px jobs on Nano Banana 2 for now. Through LaoZhang's API gateway on October 8, 2026, a 2.1 request for 512 returned a 1K image and was billed, so don't rely on an error to catch it.
- **Tokens cost more, images cost less.** The per-image price drops at every size (1K $0.0336 vs $0.067, 2K $0.0504 vs $0.101, 4K $0.113 vs $0.151). But input tokens cost $1.50 per million instead of $0.50, text and thinking output $7.50 instead of $3, and thinking now defaults to medium rather than minimal. Each 1,000 thinking tokens adds $0.0075 on 2.1 against $0.003 on Nano Banana 2. Test calls on October 8, 2026 reported 510 to 1,107 thinking tokens per image; at 1,107, a 1K image comes to about $0.0336 + $0.0083 = $0.042 before input tokens, still below Nano Banana 2's $0.067 output price on its own.
- **Sampling parameters on Vertex AI.** Google's Vertex AI docs say `seed`, `topK`, `topP`, `temperature`, and `logprobs` return an API error on 2.1, and 2.1 doesn't support Flex. Remove them from shared config. On Vertex AI, Nano Banana 2 stays supported until May 28, 2027 or later.
- **The original Nano Banana has two dates.** For `gemini-2.5-flash-image`, the Gemini API deprecations page lists shutdown on March 15, 2027 (Vertex AI says the same), while the pricing page still warns of October 2, 2026. Move it to Nano Banana 2 Lite or 2.1 rather than betting on either date.

### Requests that fail or quietly do less

| What you send | What happens | What to do instead |
| --- | --- | --- |
| `"image_size": "2k"` | Rejected: the K must be uppercase | Use `"1K"`, `"2K"`, or `"4K"` |
| `"0.5K"` or 512 on `gemini-nano-banana-2.1` | Not a size 2.1 supports | Use `"1K"`, or Nano Banana 2 for 512px |
| 2K or 4K on `gemini-3.1-flash-lite-image` | Lite supports 1K only | Switch to `gemini-nano-banana-2.1` |
| `tools=[{"type": "google_search"}]` on Lite | Grounding with Google Search isn't supported on Lite | Use Nano Banana 2.1 or Pro |
| Image search grounding for a real person | Google documents that Image Search grounding won't use web photos of real people | Supply your own reference photo |
| "Give me four variations" | The model won't always return the number of images you ask for | Make one call per image |
| `previous_interaction_id` with `store=False` | There's no stored turn to continue from | Keep the default, or resend the image |
| Any image model on a project without billing | Image models aren't available on the free tier | Set up billing in AI Studio |
| Too many or too costly requests in a short time | `429 RESOURCE_EXHAUSTED` | Back off and retry; limits rise with your usage tier |
| `gemini-3.1-flash-image` | Still works; deprecated with no shutdown date | Move to `gemini-nano-banana-2.1` |
| `gemini-2.5-flash-image` | Deprecated; Google lists shutdown as March 15, 2027 on one page and October 2, 2026 on another | Move to `gemini-3.1-flash-lite-image` or `gemini-nano-banana-2.1` |
| An `imagen-4.0-*` model | Shut down on the Gemini API on August 17, 2026 | Use `gemini-nano-banana-2.1`, Google's listed replacement |

Two more behaviors are worth knowing before you tune for cost. Thinking is always on and can't be turned off. The model may render up to two interim "thought images" to work out the composition, and those aren't charged. Through `generation_config`, `thinking_level` takes minimal, medium (the default), or high on Nano Banana 2.1, and minimal (the default) or high on Nano Banana 2 and 2 Lite; extra thinking is billed as output tokens. In one pair of test calls on 2.1, minimal didn't produce fewer thinking tokens than high, so measure before counting on the savings. For text inside an image, Google's advice is to have the model write the text first, then ask for the image that contains it.

Rate limits apply per project, not per key, and daily request quotas reset at midnight Pacific time. Linking billing puts you in Tier 1; Tier 2 takes $100 in payments plus 3 days since the first successful payment, and Tier 3 takes $1,000 plus 30 days ([rate limits](https://ai.google.dev/gemini-api/docs/rate-limits)). Your live limits are shown in AI Studio.

### What a batch of images costs

Google charges per output image, so cost = number of images × price per image, plus a token charge for your prompt, any reference images, and thinking. The Batch API halves image prices in exchange for a turnaround of up to 24 hours. For 500 images, as of October 8, 2026:

| Model and size | Standard price | 500 images | Batch price | 500 images in Batch |
| --- | --- | --- | --- | --- |
| Nano Banana 2.1, 1K | $0.0336 | $16.80 | $0.0168 | $8.40 |
| Nano Banana 2.1, 2K | $0.0504 | $25.20 | $0.0252 | $12.60 |
| Nano Banana 2.1, 4K | $0.113 | $56.50 | $0.0567 | $28.35 |
| Nano Banana 2 Lite, 1K | $0.0336 | $16.80 | $0.0168 | $8.40 |
| Nano Banana 2 (deprecated), 1K | $0.067 | $33.50 | $0.034 | $17.00 |
| Nano Banana 2 (deprecated), 2K | $0.101 | $50.50 | $0.050 | $25.00 |
| Nano Banana 2 (deprecated), 4K | $0.151 | $75.50 | $0.076 | $38.00 |
| Nano Banana Pro, 1K or 2K | $0.134 | $67.00 | $0.067 | $33.50 |
| Nano Banana Pro, 4K | $0.24 | $120.00 | $0.12 | $60.00 |

Grounding with Google Search is billed separately after 5,000 free search requests a month (shared across Gemini 3.x models), then $14 per 1,000. If you're still budgeting on Nano Banana 2, [Nano Banana 2 API Pricing](https://blog.laozhang.ai/en/posts/nano-banana-2-api-pricing-guide) covers its full rate card.

If you'd rather pay one flat price per image regardless of size or thinking level, [laozhang.ai](https://docs.laozhang.ai/en/api-capabilities/nano-banana-2-1-image) is a third-party paid API that serves the same model IDs. Its public prices as of October 7, 2026 are $0.045 per call for Nano Banana 2.1, $0.025 for Nano Banana 2 Lite, $0.055 for Nano Banana 2, $0.09 for Nano Banana Pro, and $0.02 for the legacy 2.5 model, plus $0.014 per Google Search grounding query. In eight test calls to 2.1 on October 8, 2026, from 1K up to 4K, each was charged $0.045. It is not Google, and a few conditions change the math:

*Disclosure: [laozhang.ai is our own service](https://blog.laozhang.ai/en/about#laozhang). Its prices are platform pricing, not an independent comparison.*

- Against Google's standard per-image rates, the flat price is lower for every current model and size except Nano Banana 2.1 at 1K ($0.0336 at Google before thinking tokens) and Nano Banana 2 at 0.5K. Google's Batch prices undercut it for Lite and for 1K and 2K work that can wait up to 24 hours; at 4K, the flat price stays lower even against Batch.
- Its OpenAI-compatible endpoint (`/v1/chat/completions`) can't set the aspect ratio or size and returns 1K images; its Gemini-native endpoint (`generateContent`, the legacy request shape) can.
- A response that returns HTTP 200 with no image is still billed as one call, so ask for the image explicitly in every prompt and don't auto-retry empty responses.

## Prompts that get usable images

The same prompting habits work in the app, AI Mode, Flow, and the API, drawn from Google's own advice:

- **Start with a verb, then the subject, action, and setting:** "Create an image of a cyclist crossing a rain-soaked Tokyo crosswalk at night."
- **Name the style and the camera:** "watercolor", "charcoal drawing", "photorealistic, wide-angle, low-angle shot", "macro".
- **Say what the image is for:** "a logo for a high-end, minimalist skincare brand" beats "a logo".
- **Describe what you want, not what you don't:** write "an empty, deserted street" rather than "no cars".
- **Break busy scenes into steps:** background first, then the foreground object, then the detail on top.
- **Edit instead of rerolling:** "That's great, but make the lighting a bit warmer" keeps what already works.
- **Quote exact text and check it:** put the words you need in quotes, and if 2.1 rewords them, redo the image with Pro.

Nano Banana models perform best in English and in a set of other languages that includes Spanish (Mexico), Japanese, Korean, Russian, and Simplified Chinese. For ready-made prompt structures by image type, see [How to Prompt Nano Banana](https://blog.laozhang.ai/en/posts/how-to-prompt-nano-banana).

## When to switch tools or stop

- **The Gemini app says you've reached your limit.** Check [gemini.google.com/usage](https://gemini.google.com/usage) to see whether it's the 5-hour window or the weekly limit, then wait, use AI Mode's 25 daily images, or upgrade.
- **The app runs Nano Banana 2, not 2.1.** Nano Banana 2.1 needs a Google AI plan; AI Plus at $4.99 a month in the US is the cheapest one.
- **Redo with Pro is missing.** It needs a paid plan, and it disappears for the day once your daily image quota runs out. AI Mode offers Nano Banana Pro to English-language users 18+ in the US.
- **You can't edit an uploaded photo.** Editing is limited to users 18 and older; school accounts may have other restrictions.
- **No image appears, or an image is removed.** Gemini removes images when a prompt may break Google's policies. For stuck generations and other errors, work through [Nano Banana Not Working in Gemini?](https://blog.laozhang.ai/en/posts/nano-banana-not-working-gemini-image-errors).
- **You need more than 2K.** App downloads top out at 2K with a paid plan. Use the API with `"image_size": "4K"` on Nano Banana 2.1, 2, or Pro, and see [Gemini High-Resolution Images](https://blog.laozhang.ai/en/posts/gemini-image-high-resolution-hd-prompt) to confirm the pixel size you actually got.
- **You're outside a supported country.** Each surface has its own country list. The Gemini app's list, for example, shows mainland China as Workspace-only and doesn't include Russia ([availability](https://support.google.com/gemini/answer/13575153?hl=en)). Supporting your language doesn't mean a product is available where you are.
- **Your API image call fails on a new project.** Check billing first. Image models have no free tier, so there's no free quota to wait for.

## Gemini image generation FAQ

### Is Gemini image generation free?

Yes in the Gemini app, AI Mode, and Google Flow, within limits: the app gives free accounts Nano Banana 2 with compute-based limits and no printed image count, AI Mode allows 25 images per 24 hours, and Flow gives 50 credits a day. Nano Banana 2.1 in the app needs a Google AI plan. The Gemini API has no free tier for image models; every image is billed.

### What model does Gemini use to make images?

Without a paid plan, the Gemini app uses Nano Banana 2 (`gemini-3.1-flash-image` in the API) when the model picker is on Flash or Pro, and Nano Banana 2 Lite on Flash-Lite. Google AI Plus, Pro, and Ultra get Nano Banana 2.1 (`gemini-nano-banana-2.1`) and can redo an image with Nano Banana Pro (`gemini-3-pro-image`).

### What Gemini model should I use for image generation in the API?

Use `gemini-nano-banana-2.1` for new projects: Google recommends it, and it costs $0.0336 at 1K, $0.0504 at 2K, and $0.113 at 4K. Choose Nano Banana 2 Lite for the cheapest 1K drafts and Nano Banana Pro when text inside the image has to be exact.

### How many images can I generate with Gemini a day?

In the Gemini app, Google doesn't publish a number; usage is measured as a percentage that refreshes every 5 hours up to a weekly limit. AI Mode in Google Search is the exception, with up to 25 images a day without a plan and up to 50, 100, or 1,000 on Plus, Pro, and Ultra.

### Does my old generate_content, Nano Banana 2, or Imagen code still work?

Code that calls `generate_content` with a current Nano Banana model still works; Google calls it legacy but fully supported. `gemini-3.1-flash-image` still works too, but it's deprecated with no shutdown date, so move it to `gemini-nano-banana-2.1`. Imagen 4 shut down on the Gemini API on August 17, 2026. For `gemini-2.5-flash-image`, Google's pages give two shutdown dates (March 15, 2027 and October 2, 2026), so migrate it now.

### Can Gemini make 4K images?

Through the API, yes: set `"image_size": "4K"` on Nano Banana 2.1 ($0.113 per image), Nano Banana 2 ($0.151), or Nano Banana Pro ($0.24). The Gemini app downloads at up to 2K with a Google AI plan.

## Sources

External pages this guide links to, in the order they appear. Last updated 2026-10-08.

- [gemini.google.com](https://gemini.google.com/) (gemini.google.com)
- [Gemini image page](https://gemini.google/overview/image-generation/) (gemini.google)
- [Gemini Apps Help](https://support.google.com/gemini/answer/14286560?hl=en) (support.google.com)
- [Gemini API changelog](https://ai.google.dev/gemini-api/docs/changelog) (ai.google.dev)
- [Gemini Apps limits](https://support.google.com/gemini/answer/16275805?hl=en) (support.google.com)
- [AI Mode image creation](https://support.google.com/websearch/answer/16649374?hl=en) (support.google.com)
- [Flow credits](https://support.google.com/flow/answer/16526234?hl=en) (support.google.com)
- [AI Studio plans](https://ai.google.dev/gemini-api/docs/google-ai-plans) (ai.google.dev)
- [Gemini API pricing](https://ai.google.dev/gemini-api/docs/pricing) (ai.google.dev)
- [Google subscriptions page](https://gemini.google/us/subscriptions/?hl=en) (gemini.google)
- [image generation docs](https://ai.google.dev/gemini-api/docs/image-generation) (ai.google.dev)
- [Nano Banana 2.1 model page](https://ai.google.dev/gemini-api/docs/models/gemini-nano-banana-2.1) (ai.google.dev)
- [deprecations page](https://ai.google.dev/gemini-api/docs/deprecations) (ai.google.dev)
- [gemini.google.com/usage](https://gemini.google.com/usage) (gemini.google.com)
- [Interactions overview](https://ai.google.dev/gemini-api/docs/interactions) (ai.google.dev)
- [Google AI Studio](https://aistudio.google.com/) (aistudio.google.com)
- [rate limits](https://ai.google.dev/gemini-api/docs/rate-limits) (ai.google.dev)
- [availability](https://support.google.com/gemini/answer/13575153?hl=en) (support.google.com)