Black Forest Labs Launches FLUX Upscale, a Dedicated Tool for 4K Video Upscaling

Black Forest Labs has released FLUX Upscale, a standalone tool for increasing video resolution up to native 4K. The tool is now available as its own FLUX Tool and API endpoint, separate from the company's core video and image generation models.

What It Does

According to Black Forest Labs, FLUX Upscale takes an existing video file and regenerates it at a higher target resolution, with support for both 2K and native 4K output. Rather than applying traditional interpolation-based upscaling, the tool reportedly regenerates frame content, a technique consistent with the generative approach used across the FLUX model family for image tasks like inpainting and outpainting.

The company has not disclosed the underlying architecture, parameter count, or whether FLUX Upscale shares weights with existing FLUX video or image models. No benchmark comparisons against other video upscaling methods have been published.

Availability and Pricing

FLUX Upscale is available now as a dedicated endpoint, positioning it alongside other FLUX Tools such as Fill, Canny, and Depth, which Black Forest Labs has previously released as task-specific variants built on its core FLUX.1 models. Pricing per video or per resolution tier has not yet been disclosed. No information has been released on maximum input video length, supported formats, or processing latency at 4K output.

What This Means

Video upscaling has largely been the domain of specialized tools like Topaz Video AI and traditional super-resolution models. Black Forest Labs entering this space with a generative approach signals an effort to extend its FLUX ecosystem beyond image generation and editing into video post-production workflows.

The practical value here depends heavily on details the company hasn't shared: processing cost per minute of footage, how well the generative regeneration preserves temporal consistency across frames (a common failure point for AI-based video upscaling), and whether output quality holds up against established non-generative upscalers at 4K. Until pricing and independent benchmarks surface, this should be treated as a capability announcement rather than a proven production tool. Studios and video pipelines evaluating it will want to test temporal artifacts and motion consistency directly before committing workflows to it.