【速報】新しい画像生成AIモデル FLUX.1 の紹介と試す方法について

FLUX.1 is a groundbreaking image generation AI model developed by Black Forest Labs, a venture established by co-developers of the open-source model Stable Diffusion. The model has garnered attention after securing $31 million in funding from renowned investors. It boasts high-quality image generation capabilities, accurately reflecting user prompts, and has improved upon the depiction of complex scenes and human hands, which was challenging for previous AI models. FLUX.1 is accessible to everyone, with demos available on multiple sites including Hugging Face, allowing users to easily experiment with image generation. The model comes in three versions: FLUX.1 [pro] for commercial use via API, FLUX.1 [dev] as an open-weight model for non-commercial local use with the option to inquire about commercial use, and FLUX.1 [schnell], a lightweight model released under the Apache 2.0 license for free use from personal to commercial applications. The video also discusses the pricing for API usage, performance benchmarks that place FLUX.1 ahead of major models like Stable Diffusion and Midjourney, and the copyright ownership of images generated by FLUX.1, which belongs to the user. Ethical use of the model is emphasized, with prohibitions against creating fake news or harmful content.

This paragraph delves into the performance of FLUX.1, highlighting its ability to generate high-quality images swiftly, even with minimal prompts. The model's speed is emphasized, with the capability to produce images in just four steps, drawing comparisons to lightning-fast models like SDXL Lightning. The paragraph also mentions the availability of the model file for FLUX.1 [schnell] on Hugging Face, noting the high entry barrier due to the file's large size of 23GB. The speaker shares their experience with the demo site on Hugging Face, praising the model's prompt effectiveness and high performance. They suggest that those interested in trying FLUX.1 should start with the Hugging Face demo, and provide tips on how to generate more accurate images by specifying details in the prompt. The paragraph concludes with an invitation for viewers to subscribe to the channel for more content on image generation AI and AI realism, and thanks the audience for watching.