Imagen 4
Google's Imagen 4 turns text prompts into images, with Fast, Standard, and Ultra quality tiers to match any budget.
Made with Imagen 4
Sample outputs. Open in Studio to generate your own.
Where Imagen 4 shines
Concept art
Generate quick visual concepts from a written description using the Fast tier for rapid, low-cost iteration.
Hero images
Produce high-fidelity marketing visuals with the Ultra tier in widescreen 16:9 or square 1:1 framing.
Variation sets
Request up to 8 variations from one prompt to compare options and choose the strongest result.
Strengths
- Generates images directly from a text prompt with no input or reference image needed
- Three selectable quality tiers (Fast, Standard, Ultra) let you trade speed and cost against fidelity
- Five aspect ratios available: 1:1, 16:9, 9:16, 4:3, and 3:4
- Can return up to 8 image variations from a single prompt
Trade-offs
- Text-to-image only: it does not accept a reference or input image for editing or image-to-image work
- No explicit resolution selector is exposed
- Higher quality tiers cost more, with Standard roughly 2x and Ultra roughly 3x the Fast tier per image
- Aspect-ratio choices are limited to the five offered options
At a glance
About Imagen 4
Imagen 4 is Google DeepMind's text-to-image model, available on HexGen for turning written prompts into finished images. You describe the scene you want and Imagen 4 renders it, with no input or reference image required to get started.
What sets Imagen 4 apart is its three quality tiers. Fast is the quickest, most economical option, Standard steps up the fidelity at roughly twice the per-image cost, and Ultra pushes quality highest at roughly three times the Fast price. You choose the tier that fits the job, so a quick concept sketch and a polished hero image can both come from the same model.
Every generation supports five aspect ratios, including square 1:1, widescreen 16:9, vertical 9:16, and the classic 4:3 and 3:4 frames. You can also request up to 8 variations from a single prompt, giving you a spread of options to pick from in one pass.
Starting points
Copy, tweak, and run. Good prompts get you most of the way there.
A cozy reading nook by a rain-streaked window, warm lamplight, stacked books, soft autumn palette, 4:3
Minimalist product shot of a ceramic coffee mug on a marble surface, soft studio lighting, 1:1
A lone hiker on a misty mountain ridge at sunrise, dramatic clouds, cinematic widescreen, 16:9
Pay only for what you render. 1 USD = 1,000 HGcoins. HGcoins never expire and failed runs refund automatically.
Imagen 4 vs other models
Imagen 4 is a text-to-image model with tunable quality tiers. Here is how it compares to other image models in the HexGen catalog.
| Model | Quality | Speed | Cost | Choose it when |
|---|---|---|---|---|
Imagen 4 This Google | Best | Fast | Mid cost | Pick Imagen 4 when you want Google's text-to-image quality with a Fast, Standard, or Ultra tier to dial in the speed-cost-fidelity tradeoff. |
Google | Great | Fastest | Lower cost | Choose nano-banana for a faster, lower-cost Google image option. |
ByteDance | Best | Fast | Mid cost | Choose Seedream v4.5 if you prefer ByteDance's image model for your generations. |
Related models
Frequently asked questions
Imagen 4 is Google DeepMind's text-to-image model. You give it a written prompt and it generates a matching image, with no input or reference image required.