Transparent PNGs straight out of MOSS-Image. End your prompt with <code>isolated on a transparent background</code>, decode with the RGBA VAE, and the sampler gives you an image with a real alpha channel: no chroma key, no background removal, no masking.Trained with black-and-white manga in mind (characters, creatures, props, ink effects, sound effects, speech balloons), but it also works in color.<h3 id="everything-you-need-(comfyui)">Everything you need (ComfyUI)</h3>Files, raw RGBA examples, trailer and full notes: <a target="_blank" rel="ugc" href="https://huggingface.co/paralaif/MOSS-Manga-Alpha-Image-LoRA">huggingface.co/paralaif/MOSS-Manga-Alpha-Image-LoRA</a><ol><li>This LoRA → <code>models/loras/</code></li><li>RGBA VAE <code>vae/flux2_klein_alpha_rgba_vae.safetensors</code> (in the Hugging Face repo; originally trmz/flux2-klein-alpha, Apache-2.0) → <code>models/vae/</code></li><li>Load RGBA VAE (Flux2) custom node + minimal workflow: the attached zip (also on Hugging Face). Stock ComfyUI can't load a 4-channel VAE.</li><li>MOSS-Image SFT + Qwen3.5-9B text encoder: <a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/MOSS-Image">Comfy-Org/MOSS-Image</a></li><li>8-step DMD LoRA: <code>lora.safetensors</code> from <a target="_blank" rel="ugc" href="https://huggingface.co/lodestones/MOSS-Image-SFT-8-steps-dmd">lodestones/MOSS-Image-SFT-8-steps-dmd</a>, renamed <code>moss_image_sft_8step_dmd.safetensors</code></li></ol><h3 id="tips">Tips</h3><ul><li>Always end the prompt with <code>isolated on a transparent background</code>, and describe one subject, not a place.</li><li>Say "man" or "boy" explicitly when it matters; ambiguous roles sometimes come out as women.</li><li>Semi-transparent areas (smoke, plastic umbrellas, ink sprays) can carry a faint purple tint from the RGBA VAE.</li><li>The preview images here are flattened onto colored backgrounds because Civitai doesn't show transparency. The raw RGBA PNGs are on Hugging Face.</li></ul><h3 id="training">Training</h3>MOSS-Image Base, ai-toolkit, 503 RGBA images, rank 32, AdamW 8-bit 2e-4, step 1500 of 3000, one RTX 5060 Ti 16 GB. The training images are synthetic: generated with MossAD-Demo using its built-in artist-style tags, on a flat indigo background, cut out with a chroma key and reviewed by hand. The artist tags were removed from the captions, but the LoRA may still carry some of those styles.<h3 id="license">License</h3>Apache-2.0. Credits: MOSS-Image (OpenMOSS, Apache-2.0), MossAD-Demo (RicemanT, Apache-2.0), RGBA VAE (Xavier Jara / trmz, Apache-2.0, method AlphaVAE), 8-step DMD (lodestones, Apache-2.0), ai-toolkit (Ostris, MIT).Also available for Krea 2: <a target="_blank" rel="ugc" href="https://huggingface.co/paralaif/Krea2-Manga-Alpha-Image-LoRA">paralaif/Krea2-Manga-Alpha-Image-LoRA</a>

MOSS Manga Alpha Image LoRA (transparent PNG manga)

LoRA strength 1.0. MOSS-Image SFT + 8-step DMD LoRA: 8 steps, CFG 3.5, er_sde / simple, SkipLayerGuidance on single layer 22 (0–50 %), negative <code>worst quality, bad anatomy, watermark, jpeg artifact</code>. 832×1216, 1216×832 and 1024² tested. Also works on MOSS-Image Base and MossAD.Decode with the 4-channel RGBA VAE (in the <a target="_blank" rel="ugc" href="https://huggingface.co/paralaif/MOSS-Manga-Alpha-Image-LoRA">Hugging Face repo</a>) through the Load RGBA VAE (Flux2) node from the attached zip.

Files with SHA256: 1d2a5317e5057846893350cb9350cd69b25420d9bd0de90487948a03fe9302ff