Imagen 4 is Now Supported in Flux Labs AI
You can now use Google's latest and most powerful image model, Imagen 4 in Flux Labs AI. Here's how to use it and example images to showcase its capabilities.
Google just dropped Imagen 4, its most powerful AI image model yet, during the Google I/O 2025 keynote.
Some of the key highlights in this new image model are the following:
Photorealistic images
Sharper clarity
Improved spelling and typography
Multilingual prompt support to help creators globally
API access was made immediately available to developers, so I quickly added support for Imagen 4 on Flux Labs AI.
In this post, I’ll explain what Flux Labs AI is and how you can use Imagen 4 to generate stunning images.
What is Flux Labs AI?
If you’re new to Flux Labs AI, it’s a platform that lets you generate, edit, and combine images using AI. It also lets you create user-generated content (UGC) videos to promote products or brands.
These are the key features of Flux Labs AI:
Generate AI images
Edit, combine, or remix an existing photo with AI
Upscale any image for more detail and higher resolution output
Do a virtual try-on to see how any garment looks on you
Create UGC videos to promote a product or brand on TikTok or Reels
Train an image model with your own photos
For this walkthrough, we’ll focus on the image generation feature to showcase what Imagen 4 can do.
Getting Started With Imagen 4 on Flux Labs AI
To begin, go to Flux Labs AI and open the Image Generator tool under the “AI Tools” dropdown menu.
This is what the image generator screen looks like:
Once you’re in the tool, you’ll notice a new option under the Image model dropdown: Imagen 4 — Google’s latest and most advanced image model to date.
How To Use Imagen 4 on Flux Labs AI
To use the new model, select Imagen 4 from the Image model dropdown. In case you’re wondering, the other models listed are Flux models from Black Forest Labs.
Keep in mind that supported aspect ratios depend on the selected model. For Imagen 4, these are the available options:
1:1 (Square)
16:9 (Landscape)
9:16 (Mobile)
3:4 (Portrait)
4:3
Now, open the Advanced Settings, where you can control the Safety Filter level:
Strictest: Block when low, medium, or high probability of unsafe content
Moderate: Block when medium or high probability of unsafe content
Permissive: Block when high probability of unsafe content
Text prompts provided as inputs and images (generated or uploaded) through Imagen on Vertex AI are assessed against a list of safety filters, which include “harmful categories” such as violence, sexual, or derogatory.
These safety filters aim to filter out (generated or uploaded) content that violates Google’s Acceptable Use Policy (AUP), Generative AI Prohibited Use Policy, or our AI Principles.
If the model responds to a request with an error message such as “The prompt couldn’t be submitted” or “it might violate our policies”, then the input is triggering a safety filter. If fewer images than requested are returned, then some generated output is blocked for not meeting safety requirements.
You can choose how aggressively to filter sensitive content by adjusting the safety filter parameter.
Once all the settings are set, add the description of your image in the prompt field. Here’s an example:
Prompt: A close-up captures the sheer, unadulterated joy of a Golden Retriever, its head thrust eagerly out the passenger-side window of a moving pickup truck cruising down a scenic mountainous highway. The frame is filled with the dog’s blissful expression: mouth open in a wide, happy pant, long pink tongue lolling out and catching the wind, flapping slightly. Its eyes are bright and slightly squinted against the rush of air, radiating pure delight. Golden fur ripples and streams back from its face, catching the bright sunlight which highlights the rich tones of its coat against the darker interior of the truck cab just visible at the edge of the frame. Its floppy ears are blown back dynamically by the wind, adding to the sense of exhilarating motion. Drool might glisten slightly on its jowls, a testament to its unrestrained happiness. The background outside the window is a dynamic motion blur of out-of-focus greens and greys from the passing mountain slopes, perhaps interspersed with patches of bright blue sky, effectively conveying the speed and the open-air freedom the dog is clearly savoring. The shot freezes a perfect moment of sensory delight and canine exuberance on the open road.
Awesome. The image looks hyper-realistic, and the fur looks incredibly detailed. In terms of resolution, it’s 1048 x 768 pixels and 1.1 MB in size.
Here’s the fun part: you can upscale the image to 4K with even more detail and realism. Just open the Image Upscaler tool, upload the photo, choose the Crisp Upscale model, and hit Upscale.
You’ll get a much more detailed, higher-resolution image like this one:
Here’s another example that showcases Imagen’s capability to render improved texts and typography on images.
Prompt: The Quest for the Cheesy Treasure (Retro Comic Style): Drawn with slightly pulpy art, Ben Day dots for shading, dramatic close-ups, and lurid colors (orange, deep purples, oranges). Captions use overly dramatic language. Panel one shows Mittens, an orange tabby cat, approaching a lonely, flickering vending machine in a dark breakroom late at night. Caption: “Some cravings demand EXTRAORDINARY measures!” Panel two is a close-up of the yellow glow emanating from the machine’s selection window with many cheesy puffs, illuminating Mittens’s apprehensive face. Panel three shows a cat leaping at the machine… and nothing. The machine makes a loud CLUNK and WHIRR. Panel four shows a friendly man pressing a button and a cheesy puffs bag fall out. Panel five is a dramatic shot of the triumphant tabby, bathed in lurid orange light, batting open the bag, cheese dust flying dramatically as the caption declares: “The FEAST begins!”
I love how the texts are correctly rendered on the bubble thoughts and even the “Cheesy Puffs” label on the package. This is some next-level text rendering capability!
I got curious about how the most capable Flux model handles this prompt.
I was surprised to see how inaccurate both the image composition and the text renderings are. In this example, it’s clear that Imagen 4 is much more capable than Flux 1.1 Pro Ultra.
Final Thoughts
I’m really glad that Flux Labs AI now supports Imagen 4. It’s honestly the best image model I’ve tried so far, especially when it comes to realistic textures and clean text rendering. The quality jump is obvious.
If there’s one thing that bothers me a bit, it’s Google’s safety filters. Some prompts get blocked even when they don’t seem that risky. It’s not a dealbreaker, but it does limit how creative you can get with certain descriptions.
On the technical side, getting Imagen 4 running inside Flux Labs AI didn’t take long at all. I used Cursor and the new Claude Sonnet 4 model from Anthropic. It was as easy as grabbing the LLMs.txt file from Replicate, pasting it into Cursor, and letting the AI handle the rest — frontend and backend. The whole thing took under an hour.
Anyway, give it a spin. Imagen 4 is now live on Flux Labs AI. Try it out and give me your honest feedback!
Hi there! Thanks for making it to the end of this post! My name is Jim, and I’m an AI enthusiast passionate about exploring the latest news, guides, and insights in the world of generative AI. If you’ve enjoyed this content and would like to support my work, consider becoming a paid subscriber. Your support means a lot!