Google has released Imagen 4, its newest text-to-image artificial intelligence model, through the Gemini API and Google AI Studio. The announcement was made on the official Google Developer Blog.
The company introduces two versions of the model to serve different creative needs:
- Imagen 4 serves as the flagship model for general image generation tasks, priced at $0.04 per output image.
- Imagen 4 Ultra focuses on precise prompt following and costs $0.06 per generated image.
Google claims Imagen 4 offers significant improvements in text rendering compared to previous versions. The Ultra variant is designed to produce outputs that align more closely with user instructions, according to the company.
The models are currently available in paid preview through the Gemini API, with limited free testing available in Google AI Studio. Google plans to introduce additional billing tiers and increase rate limits in the coming weeks.
All images generated by Imagen 4 include a non-visible SynthID digital watermark for transparency purposes. The company showcased the model’s capabilities with examples including comic panels, vintage postcards, and fashion photography.
Google demonstrated the model’s versatility across various artistic styles and content types. The examples included complex multi-panel comics with embedded text and detailed scenic compositions.
The company expects to make the models generally available in the coming weeks, expanding access beyond the current preview phase. Imagen 4 looks like a major competitor to ChatGPT’s viral image generating capabilities.