Google Imagen 4
Google's Imagen 4. Text-to-image with strong photorealism and improved typography support.
Google Imagen 4 is image generation AI model from Google DeepMind, priced at €0.000 per 1M input tokens with a unknown context window.
Pricing
API Integration
Use our OpenAI-compatible API to integrate Google Imagen 4 into your application.
npm install railwailimport railwail from "railwail";
const rw = railwail("YOUR_API_KEY");
const images = await rw.run("imagen-4", "A beautiful sunset over Tokyo");
console.log(images[0].url);
// Or use the image() method for full control
const res = await rw.image("imagen-4", "A cat in space", {
size: "1024x1024",
n: 1,
});
console.log(res.data[0].url);Deep dive — Google DeepMind's Google Imagen 4
Google DeepMind is the AI research organisation that resulted from the April 2023 merger of DeepMind (founded 2010 in London by Demis Hassabis, Shane Legg and Mustafa Suleyman; acquired by Google 2014) and Google Brain. It is led by CEO Demis Hassabis and CTO Koray Kavukcuoglu. Imagen is Google's text-to-image diffusion research line first introduced in May 2022 by Saharia et al. The successive versions Imagen 2 (Dec 2023), Imagen 3 (May 2024) and Imagen 4 (May 2025 at Google I/O) have been deployed through Google Cloud's Vertex AI, the Gemini app, Google Workspace and the ImageFX consumer tool. Imagen 4 represents the joint Brain-DeepMind diffusion stack and was announced as the highest-quality, fastest Imagen model to date.
Visit Google DeepMind →Imagen 4 is the May 2025 generation of Google's text-to-image diffusion family, announced at Google I/O 2025. It builds on the lineage of Imagen 1 (Saharia et al. 2022), Imagen 2 (Dec 2023) and Imagen 3 (Hayes et al. 2024). The original Imagen used a frozen T5-XXL language model for text conditioning followed by a pixel-space cascaded diffusion super-resolution pipeline (64 -> 256 -> 1024). Imagen 3 and 4 are widely believed to have transitioned to a latent diffusion approach with a strong VLM-based text encoder (likely a Gemini text encoder), but Google has not disclosed full architecture details. Imagen 4 is reported to deliver substantially better photorealism, lighting and texture rendering than Imagen 3, with improved long-prompt understanding and text-in-image fidelity. The model is available exclusively via Google Cloud Vertex AI, Gemini and Workspace integrations. Outputs ship with SynthID invisible watermarks.
- Parameters
- Undisclosed (estimated multi-billion)
- Context
- 480 tokens
- Photorealistic image generation up to 2K resolution
- Strong long-prompt understanding (480 tokens)
- Improved in-image typography vs. Imagen 3
- SynthID invisible watermark on every output
- Integration with Gemini app and Workspace
- Multilingual prompt support
- Safety filters tuned for advertising and brand-safe use
- Best for: enterprise creative workflows, ads, presentations, knowledge-graph-grounded imagery.
Trained on a large licensed and publicly available image-text dataset filtered for safety, bias and IP. Exact composition is not disclosed. Google states training included substantial multilingual and culturally diverse data.
License: Proprietary commercial license via Google Cloud Vertex AI, Gemini API and Google Workspace.
Known limitations
- Vertex AI only — no public weights
- Higher refusal rate for named persons and brands than open models
- Limited control compared to ControlNet pipelines
- Pricing higher than Flux schnell tier
Frequently asked questions
Related Models
View all Image GenerationFlux 1.1 Pro Ultra
FLUX 1.1 Pro in ultra mode. Up to 4 megapixel images with raw mode for photorealism.
Flux Dev
Black Forest Labs' development model. Fast, high-quality image generation with LoRA support.
Google Imagen 4 Ultra
Premium Imagen 4 tier. Highest fidelity, prompt adherence and typography quality from Google.
Ideogram 3.0
Ideogram's flagship text-to-image model with industry-leading text rendering and prompt adherence.
Start using Google Imagen 4 today
Get started with free credits. No credit card required. Access Google Imagen 4 and 100+ other models through a single API.