Flux Identity Adjuster V2 node kills waxy skin in flux.2
New frequency filtering and overdrive boost photorealism for character-consistent images.
The Flux Identity Adjuster V2 is a community-built node (by Reddit user Stock_Mycologist1104) designed to improve character consistency and reduce the waxy skin artifact often seen in images generated with the flux.2 klein 9B FP8 distilled model. Building on the original Flux ID Adjuster, V2 introduces two core enhancements: frequency filtering and apply overdrive. The frequency filtering drop-down allows users to extract high and low frequency data from an input image and route them separately to the S and D blocks of the diffusion model, giving finer control over texture and sharpness.
The apply overdrive option amplifies block residuals during generation, boosting detail, realism, and overall punch. This feature was ported from a custom overdrive node previously created for photorealism. The author generated comparison sets of images at 1MP resolution (without the node, with the node, and with overdrive active) using random prompts from various sites. The resulting images clearly demonstrate mitigated waxy skin while maintaining character identity. The node is available on GitHub with a full workflow attached.
- Frequency filtering extracts and routes high/low frequency data separately to S and D blocks.
- Apply overdrive amplifies block residuals to enhance detail and realism.
- Tested on flux.2 klein 9B FP8 at 1MP; eliminates waxy skin effect while preserving character consistency.
Why It Matters
This free node gives AI image creators practical control over photorealism, solving a common artifact in popular flux models.