Image generation (experimental)
Blog post from Ollama
Ollama has introduced image generation capabilities on macOS, with plans to expand to Windows and Linux. Users can generate images by running "ollama run x/z-image-turbo" along with a prompt, and the images save to their current directory, with terminals like Ghostty and iTerm2 supporting inline previews. The Z-Image Turbo model, developed by Alibaba's Tongyi Lab, features 6 billion parameters and excels in creating photorealistic images with bilingual text rendering in English and Chinese. It is available under Apache 2.0 for commercial use. Another model, FLUX.2 Klein from Black Forest Labs, is optimized for text rendering in images, available in 4B and 9B parameter sizes, with the former open for commercial use under Apache 2.0. Users can customize image generation by adjusting parameters such as image size, number of steps, random seed, and negative prompts. Future developments include extending support to Windows and Linux, adding new image generation models, and incorporating image editing capabilities.