Google Imagen 4 guide: capabilities, current access, and alternatives
Quick answer
Imagen 4 is Google's text-to-image model family, with Fast, standard, and Ultra variants. It generates still images from written prompts. Access now depends on the provider: Imagen 4 shut down in the Gemini Developer API on August 17, 2026, while Google Cloud still documents the model family. Check the exact endpoint before following an older tutorial.
A quiet glass observatory above a cloud layer at blue hour, warm interior lights, one telescope, cinematic realistic lighting, balanced composition, fine architectural detail, no people, text, or logos.
Compare an image brief in the browser
Generate a representative prompt, record the current model and settings, and inspect text, faces, products, and fine detail at full resolution.
Try AI Image GeneratorWhat can you use Imagen 4 for?
Use it to explore illustrations, product scenes, backgrounds, and other still-image concepts. Describe the subject, action, composition, lighting, and output shape. Review any text or product detail before treating the image as finished artwork. A model's ability to generate text does not guarantee correct spelling or faithful logos.
Google's Imagen 4 documentation explains supported features and model settings. Its Gemini API deprecation schedule explains the closed Developer API route. References checked September 9, 2026.
How to get started
- Choose a provider that currently offers the model and confirm access in your account.
- Select the model variant, supported aspect ratio, and number of outputs.
- Write one clear prompt that describes the important visual requirements.
- Generate a small set and inspect every candidate at the intended display size.
- Revise one requirement at a time, then export the accepted image.
For example: A matte blue ceramic mug on a pale wood desk, soft window light from the left, three-quarter view, empty space on the right for a headline, square composition. Add final advertising text in a design editor when exact typography is essential.
Which variant should you choose?
Fast, standard, and Ultra have different prices. Compare their accepted results on your own prompt rather than assuming the most expensive version always wins. Google Cloud's published generation rates are $0.02, $0.04, and $0.06 per image respectively. See the pricing and API guide for the source, provider distinction, and a worked budget.
Can Imagen 4 make video?
Imagen 4 is an image-generation model. To add motion to an accepted image, use a separate image-to-video workflow. For example, upload it to Magic Hour image-to-video and describe a restrained camera move or subject action. Check the resulting clip for unwanted changes to the source.
What should you check before publishing?
Inspect text, hands, product shapes, faces, and any factual or branded details. Confirm that your use of source assets and the generated output meets the provider's terms. Keep the prompt and model name with your project so you can understand what produced the result.
Frequently asked questions
Use it to explore illustrations, product scenes, backgrounds, and other still-image concepts. Describe the subject, action, composition, lighting, and output shape. Review any text or product detail before treating the image as finished artwork. A model's ability to generate text does not guarantee correct spelling or faithful logos.
Google's Imagen 4 documentation explains supported features and model settings. Its Gemini API deprecation schedule explains the closed Developer API route. References checked September 9, 2026.
Fast, standard, and Ultra have different prices. Compare their accepted results on your own prompt rather than assuming the most expensive version always wins. Google Cloud's published generation rates are $0.02, $0.04, and $0.06 per image respectively. See the pricing and API guide for the source, provider distinction, and a worked budget.
Imagen 4 is an image-generation model. To add motion to an accepted image, use a separate image-to-video workflow. For example, upload it to Magic Hour image-to-video and describe a restrained camera move or subject action. Check the resulting clip for unwanted changes to the source.
Inspect text, hands, product shapes, faces, and any factual or branded details. Confirm that your use of source assets and the generated output meets the provider's terms. Keep the prompt and model name with your project so you can understand what produced the result.
Runbo Li is the Co-founder and CEO of Magic Hour, where he builds AI video and image tools for content creation. He is a Y Combinator W24 founder and former Data Scientist at Meta, where he worked on 0-1 consumer social products in New Product Experimentation. He writes about AI video generation, AI image creation, creative workflows, and creator tools.
