News from the AI & ML world
DeeperML
@learn.aisingapore.org
//
Black Forest Labs, founded by alumni of the original Stable Diffusion project, has launched FLUX.1 Kontext, a new suite of generative flow matching models designed for in-context image generation and editing. Released on May 29, 2025, FLUX.1 Kontext allows users to generate and edit images by leveraging both text and reference images as input. This differs from traditional text-to-image models, enabling more precise and iterative control over the creative process. The release includes two models, Kontext [pro] and Kontext [max], accompanied by a web Playground that runs the same API, providing developers and creatives with tools for seamless edits and new scene creation.
Kontext [pro] and [max] build upon a 12 B-parameter rectified-flow transformer, optimized for rapid sampling. Internal tests boast up to 8x lower latency compared to FLUX 1.1 Pro, along with top scores on the new KontextBench for text editing and character preservation. Key features include local edits, style transfer, on-image text rewrite, and stable character identity. The “max” variant further enhances typography quality and prompt following. These models are currently accessible through the BFL API and partner services like KreaAI, Freepik, Lightricks, Leonardo, Replicate, FAL, Runware, and Together, with an open-weight 12 B Kontext [dev] planned for wider release on Hugging Face after a safety review.
Developers utilizing the Replicate endpoint have reported crisper edits at a lower cost compared to OpenAI’s gpt-image-1, praising its ability to maintain consistent subjects throughout multi-step workflows. Black Forest Labs aims to deepen its hybrid strategy by pairing commercial high-speed endpoints with research weights that the community can fine-tune. The company emphasized that FLUX.1 Kontext offers character consistency, preserving elements across scenes; local editing, targeting specific parts without affecting the rest; style reference, generating scenes in existing styles; and minimal latency. The models have been made available on platforms such as KreaAI, Freepik, Lightricks, OpenArt, and LeonardoAI, to allow enterprise creative teams and other developers to edit images with precision and at a faster pace.
References :
Kontext [pro] and [max] build upon a 12 B-parameter rectified-flow transformer, optimized for rapid sampling. Internal tests boast up to 8x lower latency compared to FLUX 1.1 Pro, along with top scores on the new KontextBench for text editing and character preservation. Key features include local edits, style transfer, on-image text rewrite, and stable character identity. The “max” variant further enhances typography quality and prompt following. These models are currently accessible through the BFL API and partner services like KreaAI, Freepik, Lightricks, Leonardo, Replicate, FAL, Runware, and Together, with an open-weight 12 B Kontext [dev] planned for wider release on Hugging Face after a safety review.
Developers utilizing the Replicate endpoint have reported crisper edits at a lower cost compared to OpenAI’s gpt-image-1, praising its ability to maintain consistent subjects throughout multi-step workflows. Black Forest Labs aims to deepen its hybrid strategy by pairing commercial high-speed endpoints with research weights that the community can fine-tune. The company emphasized that FLUX.1 Kontext offers character consistency, preserving elements across scenes; local editing, targeting specific parts without affecting the rest; style reference, generating scenes in existing styles; and minimal latency. The models have been made available on platforms such as KreaAI, Freepik, Lightricks, OpenArt, and LeonardoAI, to allow enterprise creative teams and other developers to edit images with precision and at a faster pace.
Share:
References :
- AI Talent Development: Fuel your creativity with new generative media models and tools
Classification:
- HashTags: #GenerativeAI #AICreativity #MediaCreation
- Company: AI Labs
- Target: Artists, content creators
- Product: Veo 3, Imagen 4, FLUX 1 Kontext
- Feature: Image Editing
- Type: AI
- Severity: Informative