Imagen (text-to-image model)
| Imagen | |
|---|---|
| An image generated with Imagen 4. Partial prompt: Softly illuminated afternoon valley with meandering river | |
| Developer | Google DeepMind |
| Initial release | May 2022; 3 years ago (2022-05) |
| Stable release | Imagen 4
/ 20 May 2025; 6 months ago (2025年05月20日) |
| Type | Text-to-image model |
| Website | Imagen website |
.webp){kind=link}
| Part of a series on |
| Artificial intelligence (AI) |
|---|
|
Glossary |
_artificial_intelligence_icon.png){kind=link}
Imagen is a series of text-to-image models developed by Google DeepMind. They were developed by Google Brain until the company's merger with DeepMind in April 2023.[1] Imagen is primarily used to generate images from text prompts, similar to Stability AI's Stable Diffusion, OpenAI's DALL-E, or Midjourney.
The original version of the model was first discussed in a paper from May 2022.[2] The tool produces high-quality images and is available to all users with a Google account through services including Gemini, ImageFX, and Vertex AI.[3]
History
[edit ]Imagen's original version was first presented in a paper published in May 2022. It featured the ability to generate high-fidelity images from natural language.[2] The second version, Imagen 2 was released in December 2023.[4] The standout feature was text and logo generation.[5] Imagen 3 was released in August 2024.[6] Google claims that the newest version provides better detail and lighting on generated images.[7] On 20 May 2025 at Google I/O 2025 the company released an improved model, Imagen 4.[8]
Technology
[edit ]Imagen uses two key technologies. The first is the use of transformer-based large language models, notably T5, to understand text and subsequently encode text for image synthesis. The second is the use of cascaded diffusion models providing high-fidelity image generation. Imagen generates image in three stages, starting from a base of 64x64, then upsampled to 256x256 and 1024x1024.[2] Imagen 4 generates image up to 2k. [9]
Capabilities
[edit ]Imagen can generate photorealistic images from text prompts.[3] It can also create various styles, such as cinematic, 35mm film, illustration, and surreal. Like most text-to-image generative AI models, Imagen has difficulty rendering human fingers, text, ambigrams and other forms of typography.
The model can generate images in five aspect ratios, namely 9:16, 3:4, 1:1, 4:3, and 16:9. Imagen can also refine already generated images by editing existing text prompts.[7]
See also
[edit ]References
[edit ]- ^ Roth, Emma; Peters, Jay (April 20, 2023). "Google's big AI push will combine Brain and DeepMind into one team". The Verge . Archived from the original on April 20, 2023. Retrieved March 18, 2025.
- ^ a b c Saharia, Chitwan; Chan, William; Saxena, Saurabh; Li, Lala; Whang, Jay; Denton, Emily; Seyed Kamyar Seyed Ghasemipour; Burcu Karagol Ayan; Sara Mahdavi, S.; Rapha Gontijo Lopes; Salimans, Tim; Ho, Jonathan; David J Fleet; Norouzi, Mohammad (2022). "Photorealistic Text-to-Image Diffusion Models with Deep Language Understanding". arXiv:2205.11487 [cs.CV].
- ^ a b Peterson, Jake (2024年08月16日). "Anyone With a Google Account Can Try Google's Latest AI Image Generator Right Now". Lifehacker. Retrieved 2025年03月18日.
- ^ "Imagen 2 - our most advanced text-to-image technology". Google DeepMind. 2025年03月12日. Retrieved 2025年03月18日.
- ^ Wiggers, Kyle (2023年12月13日). "Google debuts Imagen 2 with text and logo generation". TechCrunch. Retrieved 2025年03月18日.
- ^ Schoon, Ben (2024年08月16日). "Google opens access to Imagen 3, its latest model for AI image generation". 9to5Google. Archived from the original on 2024年08月18日. Retrieved 2025年03月18日.
- ^ a b Christian Rowlands (2025年02月26日). "Some of the most realistic AI images you'll see were created with this free tool". TechRadar. Retrieved 2025年03月18日.
- ^ Kyle Wiggers (2025年05月20日). "Imagen 4 is Google's newest AI image generator". techcrunch.com. Retrieved 2025年03月18日.
- ^ "Imagen". Google DeepMind. Retrieved 2025年10月28日.
