The Ultimate Google Flow Guide: Unlimited Free AI Images With Consistent Characters

Most people use Gemini’s chat interface to generate images—and then wonder why the results feel random, inconsistent, or low quality. There’s a much better way to work with Google’s image models: Google Flow. It gives you a visual workspace, precise control over edits, consistent characters, and even complex multi-image compositions, all for free with Nano Banana 2.

Below is a complete walkthrough of how to use Google Flow to create high-quality, watermark-free images, build scenes, and keep characters consistent across shots.

What Is Google Flow and Why Use It?

Google Flow is a visual workspace built for working with Google’s image models (like Nano Banana and Nano Banana Pro). Instead of chatting back and forth with an AI, you work on a canvas where every generation, edit, and variation is clearly tracked.

Key advantages over basic chat-based image generation:

• No watermarks on generated images
• Unlimited use of Nano Banana 2 on the free plan
• Better control over aspect ratios and batch generation
• Editable history with a timeline of every change
• Consistent characters and styles across multiple shots

If you’re into filmmaking or visual storytelling, this workflow pairs nicely with other creator-focused tools like Google’s Fabula screenwriting system, which we cover in detail in this guide to Fabula and Nvidia’s Image-to-City AI.

Getting Comfortable With the Google Flow Interface

When you open Google Flow, you’ll see a blank canvas where you can type prompts, name your project, and manage all your assets on the side.

To generate your first image:

1. Type a prompt like “a woman in red armor” into the canvas box.
2. Hit send—Flow immediately starts generating the image without extra confirmation steps.
3. You’ll see the prompt and the resulting image appear in your workspace.

From there, you can control some important settings:

• Aspect ratio: Choose 9:16 for vertical content (TikTok, Shorts) or 16:9 for traditional YouTube and widescreen shots.
• Batch size: Generate up to four images at once so you don’t have to wait on a single result each time.
• Model selection: Use Nano Banana Pro if you have access, or Nano Banana 2 for unlimited free generations.

All of this happens in one clean UI designed for visual work rather than chat.

How Editing Works in Flow (And Why Context Matters)

One of the first things to understand about Google Flow is how context works. The main prompt box at the bottom of a fresh canvas is stateless—each time you type there, Flow treats it as a brand-new request with no memory of previous images.

That means if you type “make the woman turn to the right” in a fresh box, Flow has no idea which woman you mean.

Editing a Specific Image

To edit an existing image correctly:

1. Click on the image you want to modify.
2. This opens a dedicated context window for that specific image.
3. Now type your edit prompt, such as “turn the woman to the right”.
4. Hit enter, and Flow will generate a new version based on that exact image.

You’ll see a timeline or stack on the right side: the original image, then each edit as a new step. You can click back to any earlier version and branch off with new edits from there.

Targeted Area Edits (Local Changes)

Flow also supports localized edits, which is where it really pulls ahead of simple chat image tools.

For example, to add snow to distant mountains:

1. Open the image in its context window.
2. Use the selection tool to draw a red circle around the mountains.
3. Type a prompt like “Make the mountains in the red circle be covered in snow.”
4. Hit enter and let Flow regenerate the image.

The result: only the selected area changes, while the rest of the image stays intact. This makes it easy to refine scenes without starting from scratch.

Keeping Characters and Styles Consistent Across Scenes

One of the biggest challenges in AI image generation is consistency—especially if you’re telling a story, designing a comic, or building a visual world. Flow’s “Add to prompt” feature solves this elegantly.

Creating the Next Shot With the Same Character

Let’s say you’ve created a great shot of a woman in red armor on a cliff, and now you want a new angle where she’s walking toward the edge.

Here’s how to keep her look and style consistent:

1. Click the final image you like.
2. Open the three-dot menu and choose “Add to prompt.”
3. This attaches the image as a reference to your next generation.
4. In the prompt box, write something like “The woman in red armor walking towards the edge.”
5. Set batch size to four if you want multiple options, then generate.

Because the reference image is attached, Flow preserves the character’s design and the anime art style across the new shots. You can then pick your favorite result and continue the process.

Designing a Second Character in the Same Style

Now imagine you want to introduce a second character—a man in blue armor—who should match the visual style of your existing world.

To do this:

1. Take your original front-facing image of the woman in red armor.
2. Click “Add to prompt” to attach it as a style reference.
3. In the prompt, write something like “Character sheet for a man in blue armor.”
4. Generate a batch of four character sheets.

Flow uses the attached image as a style guide, so the new character will naturally fit into the same universe. You can then pick your favorite design—say, a knight you decide to call Sir Valerius.

Note: Flow references attached images in order. The first attached image is image 1, the second is image 2, and so on, up to around 12 images. This ordering becomes important for more advanced prompting.

Combining Multiple Characters and a Setting With JSON Prompts

When you want to merge several references—like a background setting plus two different character sheets—simple text prompts can start to break down. This is where JSON-style prompting becomes extremely useful in Google’s ecosystem.

The idea is to:

• Feed a helper GPT or prompt builder your reference images and a short description of what you want.
• Let it output a structured JSON prompt that clearly defines how each image should be used.
• Paste that JSON into a fresh Flow chat, attach the same images in the same order, and generate your final composition.

Step-by-Step: Two Characters on a Cliff

Here’s how you might pose both the woman in red armor and Sir Valerius on the same cliff, looking over the ocean:

1. Collect your three key images:
• The setting: your favorite shot of the cliff and ocean.
• Character 1: the woman in red armor.
• Character 2: Sir Valerius in blue armor.

2. Use a prompt-building GPT:
• Provide the three images.
• Write a short instruction like: “The woman in red armor standing next to the man in blue armor, both at the edge of the cliff, viewed from the side.”
• The GPT returns a JSON prompt that references image1, image2, and image3 clearly.

3. Back in Google Flow:
• Open a fresh chat window.
• Paste the JSON prompt into the text box.
• Attach the images in the exact same order you used with the GPT:
– Image 1: the setting (cliff and ocean).
– Image 2: the woman in red armor.
– Image 3: Sir Valerius.
• Set batch size (e.g., four images) and generate.

The result: multiple high-quality options where both characters are posed together in the same scene, with consistent style and composition that matches your original world.

JSON prompting might look intimidating at first, but it’s one of the most powerful ways to control complex AI image workflows—especially when you’re mixing multiple characters, styles, or reference shots.

Why This Workflow Matters for Creators

Using Google Flow instead of a basic chat interface unlocks a more professional, film-friendly workflow:

• Storyboard-level control: Build sequences shot by shot with consistent characters and angles.
• High resolution: Download images at 2K or 4K (depending on your plan).
• No watermarks: Clean outputs ready for pitch decks, animatics, or concept art.
• Scalable workflow: Batch generations and structured prompts help you move fast.

If you’re serious about visual storytelling, this approach pairs well with other free, high-output creative tools. For example, you can combine Flow-generated images with the free video tools covered in our guide to creating unlimited AI videos with no watermarks to build complete animatics or motion pieces.

Once you get comfortable with image context, reference ordering, and JSON-style prompts, Google Flow becomes a powerful, free alternative to many paid image tools—especially if you need consistent characters and cinematic scenes.