Transparent Backgrounds From Any LoRA: Render on White, Re-render on Black
emilrueh · reddit · 2026-10-14
A developer using FLUX.2 klein 4B with a style LoRA shares a reliable transparent-background workflow when LayerDiffuse, alpha-capable hosted models, and background removers all fail: generate on white, feed it back as a reference to re-render the same subject on black (seed alone isn't enough—11-29% of pixels moved on rerun), then compute alpha = 1 - (white - black)/255 per classic triangulation matting (Smith & Blinn 1996). Enclosed gaps remain a known failure handled by a constrained segmentation pass.
More from Multimodal
- Lightricks' APEX pairs hierarchical latents with async denoising for extreme 64-128x diffusion compression — Lightricks · 2026-10-14
- China's AI short-drama flood backfired: viewer fatigue killed revenue and dragged down live-action too — TuhinChakr · 2026-10-14
- One prompt, 8.5-minute sitcom: fully local video pipeline runs on a single RTX 4090 — teachersecret · 2026-10-14
- Should Lighting Be Baked Into AI Environment Images or Added at Final Video Generation? — robertwellesley · 2026-10-14
- User Trains and Open-Sources a Zenless Zone Zero Chibi Style LoRA for Qwen Image 2.1 — Unusual-Customer713 · 2026-10-14
- kimo v3: Open-Source GUI-less Video Editor Where an Agent Cuts a Full 2-Minute Video via MCP — LieTop7965 · 2026-10-14