This week, NVIDIA spotlights its latest collaboration with Black Forest Labs, unveiling FLUX.1 Kontext [dev], an advanced AI model that unifies image generation and editing into a single, seamless experience—now supercharged with TensorRT for RTX GPUs.
Unlike traditional tools requiring complex setups or multiple models, FLUX.1 Kontext [dev] allows creators to edit images intuitively using natural language and reference images, with no fine-tuning required. Thanks to NVIDIA’s TensorRT and quantization optimizations, the model now runs with up to 2x faster inference and lower VRAM use, making powerful generative AI tools more accessible than ever.
Key Features:
- Real-time editing on RTX-powered systems
- Character consistency across scenes
- Localized changes without affecting the whole image
- Style transfer using reference images
- Available as open weights on Hugging Face (Torch + TensorRT), and integrated into tools like ComfyUI
An online playground is live for testing, with an NVIDIA NIM microservice version coming in August. Developers can expect sample code and a new DemoDiffusion repo later this month.
Bonus: More from NVIDIA
NVIDIA now supports Google’s Gemma 3n multimodal model on RTX and Jetson, and is hosting the Project G-Assist Plug-In Hackathon through July 16. Save the date for the G-Assist Plug-In webinar on Wednesday, July 9, from 10-11 a.m. PT, to learn more about Project G-Assist capabilities and fundamentals, and to participate in a live Q&A session.
See the blog here: https://blogs.nvidia.com/blog/rtx-ai-garage-flux-kontext-nim-tensorrt/