Latest news with #BF16

Techday NZ

NVIDIA & Black Forest Labs boost AI image editing with FLUX.1

NVIDIA has partnered with Black Forest Labs to optimise the FLUX.1 Kontext image generation model for RTX GPUs using TensorRT. Black Forest Labs has developed the FLUX.1 Kontext model to further simplify the process of guiding and refining AI-generated images. Unlike traditional workflows that combine multiple models and rely on ControlNets for fine-tuning, FLUX.1 Kontext offers a single solution for both generating and editing images through natural language. This approach enables users to start with a reference image and direct edits using simple language prompts, eliminating complex multi-model workflows. The model handles both text and image inputs, allowing users to reference a visual concept and guide its development in a more coherent and intuitive manner. Model capabilities The FLUX.1 Kontext model offers several core features, including character consistency, localised editing, style transfer, and real-time performance. Black Forest Labs describes the key capabilities as follows: Character Consistency: Preserve unique traits across multiple scenes and angles. Localised Editing: Modify specific elements without altering the rest of the image. Style Transfer: Apply the look and feel of a reference image to new scenes. Real-Time Performance: Low-latency generation supports fast iteration and feedback. The goal is to enable coherent, high-quality edits that remain faithful to the original concepts. By providing both natural language and image-based editing options, FLUX.1 Kontext aims to make the refining process more accessible to a broader range of users, without the need for technical expertise or additional models. Performance optimisations NVIDIA collaborated with Black Forest Labs to optimise FLUX.1 Kontext for RTX GPUs using the TensorRT software development kit. This includes quantising the model to reduce VRAM requirements and improve accessibility for users running it locally. According to NVIDIA, these changes deliver more than twice the acceleration compared to running the original BF16 model with PyTorch, allowing for lower latency and faster iteration times in real-time editing workflows. As described by Black Forest Labs, the optimisation was designed to open up access to the benefits of high-fidelity AI image editing to a larger audience: "To further streamline workflows and broaden accessibility, NVIDIA and Black Forest Labs collaborated to quantise the model - reducing the VRAM requirements so more people can run it locally - and optimised it with TensorRT to double its performance. Thanks to TensorRT - a framework to access the Tensor Cores in NVIDIA RTX GPUs for maximum performance - users gain access to over 2x acceleration compared with running the original BF16 model with PyTorch." Availability and developer support FLUX.1 Kontext [dev] is now available for download in both Torch and TensorRT variants on the Hugging Face platform. Users can run the Torch models in ComfyUI, and Black Forest Labs has also made an online playground available for broader experimentation. For developers and advanced users, NVIDIA is preparing sample code to support the integration of TensorRT pipelines, with additional resources expected to be released later this month. The release of FLUX.1 Kontext follows a period of increased interest in adaptable, user-friendly AI image generation solutions. By combining natural language guidance, visual references, and enhanced GPU optimisation, the companies aim to further reduce barriers to AI-powered image editing for both hobbyists and professionals. Follow us on: Share on: