Adapted by:
Javier Pastor
Senior WriterComputer scientist turned tech journalist. I've written about almost everything related to technology, but I specialize in hardware, operating systems and cryptocurrencies. I like writing about tech so much that I do it both for Xataka and Incognitosis, my personal blog.
Karen Alfaro
WriterCommunications professional with a decade of experience as a copywriter, proofreader, and editor. As a travel and science journalist, I've collaborated with several print and digital outlets around the world. I'm passionate about culture, music, food, history, and innovative technologies.
AI’s ability to generate text, images, and video has proven impressive. Until now, however, these systems couldn’t restore old photos to perfection.
That just changed with Flux.1 Kontext, a family of generative AI models with a special feature. Instead of relying on the traditional text-to-image model, Flux.1 Kontext generates in-context images. A reference image and text prompt combine to create a more powerful result.
Black Forest Labs, the startup behind the models, highlights the range of possibilities. The system can preserve context—landscapes, faces, objects—while allowing users to modify other parts of the image. By combining reference images with detailed prompts, users can achieve remarkably precise results.
While Black Forest Labs didn’t focus on it, the same concept applies to a timeless need: restoring old photos. Upload any image damaged by time or other factors, and Flux performs a remarkable restoration in moments.
An old (I won’t say how old) photo of mine. From a distance, the result looks spectacular—credible, even. But when I zoom in, things start to feel off. My face looks a bit more elongated, squarer. The nose and forehead seem longer. It’s me, but not quite me.
You can test the tool on Black Forest’s playground or services like Fal.ai. We tested it on Freepik, a generative AI platform that has offered Black Forest Labs models since the beginning. Freepik now supports Flux.1 Kontext in two modes: Pro and Max.
The results are striking. The system can restore old photos with vivid new color and greater detail than the originals. But with portraits, results can sometimes appear unnatural or overly retouched. Facial features may look altered, creating a slightly unsettling effect.
This can be counterproductive. We used a simple prompt for our test: “Restore and colorize this image. Remove any scratches or imperfections; maintain fidelity.” More precise prompts can help fine-tune the output.
Of course, the technology has limits. When zooming into images, users may notice “hallucinations”—small details invented by AI models. In the famous Lunch atop a Skyscraper photo, for instance, the original lacks fine detail. When the AI attempts to restore it, it fabricates minor elements. Still, the overall result looks fantastic from a distance.
While Flux.1 Kontext doesn’t significantly enhance the image’s definition, it prioritizes fidelity—restoring color and repairing all the cracks in the original photo. Source: Reddit.
The examples speak for themselves. This model can deliver spectacular photo restoration, as seen in the family photo shared on Reddit.
To get a sense of the cost: each restoration using Flux Pro on Freepik costs 100 credits. With a Premium monthly subscription for $20, users get 18,000 credits.
Other Black Forest Labs models are available on GitHub and can run locally if you have a capable GPU. However, Flux.1 Kontext isn’t included in that catalog—for now. It remains unclear whether the company will release it there. In the meantime, users can experiment with it for free on the platforms mentioned above.
It’s clearly an outstanding tool for photo restoration and more.
Image | Xataka On with Flux.1 Kontext
View 0 comments