Zhenzhen's AI Radar · Flux.3 Klein · 2026-11-09 Morning Brief
This issue features 1 new versions: Strongly Recommended 0 · Recommended 1 · Worth Watching 0 · Included Only 0
Editor's Note This issue reviews 1 new versions of Flux.3 Klein, rated A. Top pick: Foreground Extraction LoRA. It directly generates RGBA PNGs with true transparency channels from the subject in existing RGB images. It can automatically extract the subject, or you can specify the object to extract via prompts. It relies on the author's custom nodes and a 4-channel RGBA VAE. All 2 example images are age-restricted; check the workflow included in the node repository before use.
Recommended (1)
[A] FLUX 3 Image Klein Native RGBA Foreground Extraction & Matting LoRA · v1.0
LoRA · New Model · Author denrakeiw · Released 11-08 21:17 · Downloads 33
Foreground extraction LoRA for Flux.3 Klein: Input a standard RGB image, combined with a 4-channel RGBA VAE, to directly output an RGBA PNG with a transparent background. Semi-transparent details like hair, glass, and glowing edges are reconstructed in latent space, rather than applying a 2D mask.
Images
Reasons for Recommendation
- New model; serves e-commerce cutouts, sticker/asset creation, and subject separation before compositing. It can automatically extract the subject, or you can specify which object to extract in the prompt.
- The author claims it differs from segmentation masks like rembg and BiRefNet, capable of removing background color spill on hair and semi-transparent surfaces.
- Full parameters provided: LoRA strength 1.0, euler / simple, 20 steps, FluxGuidance 3.5, resolution 1024x1024 or based on input image aspect ratio; based on FLUX 3 Image Klein 5.6B.
- License allows commercial use and sale of outputs; 33 downloads and 7 likes within 12 hours of release (indicating attention only).
How to Use
- Use Load Flux 2/3 RGBA VAE from DenRakEiw_Nodes to load flux2-klein-alpha-vae, use Flux 2/3 RGBA VAE Encode to encode the input image, and connect it to the positive conditioning via ReferenceLatent; text encoder qwen_3_5_4b (CLIPLoader type flux3).
- Automatic subject extraction: Prompt "Extract the main foreground subject, RGBA transparent background, isolated foreground object"; Specify object extraction: Prompt "Extract foreground object, RGBA transparent background:" followed by the object description.
Risks and Limitations
- Depends on the author's custom node package and a separately downloaded 4-channel RGBA VAE; standard Flux.3 Klein workflows cannot produce transparency channels by direct loading.
- All 2 example images are age-restricted, so edge extraction quality is not visible; "clean edges, no color spill" is the author's claim.
- This entry has not been tested.
Trigger Words: Extract the main foreground subject, RGBA transparent background, isolated foreground object
Data: Civitai Public API (Base Model Flux.3 Klein), inclusion window 11-08 20:31 → 11-09 09:18 (Beijing Time). Recommendation reasons reviewed by editors; download counts indicate attention only.