Text-to-image with a real alpha channel for MOSS-Image. End your prompt with isolated on a transparent background, decode with the RGBA VAE from trmz/flux2-klein-alpha, and get a transparent PNG straight from the sampler: no chroma key, no background removal, and partial transparency where it belongs (glass, fairy wings, gems, the Eiffel Tower's lattice).

⚠️ Proof of concept: trained on only 39 images in ~40 minutes on one 16 GB GPU. Not perfect (expect occasional halos or semi-transparent patches), but it shows MOSS-Image can learn alpha with very little data.

Required (ComfyUI)

1. MOSS-Image SFT + Qwen3.5-9B text encoder (Comfy-Org/MOSS-Image)

2. RGBA VAE: vae/diffusion_pytorch_model.safetensors from trmz/flux2-klein-alpha (Apache-2.0)

3. The Load RGBA VAE (Flux2) custom node and the workflow, from paralaif/MOSS-Image-Alpha-LoRA (folder comfyui/). Stock ComfyUI can't load a 4-channel VAE.

4. Optional: an 8-step DMD LoRA for MOSS-Image SFT (e.g. lodestones/MOSS-Image-SFT-8-steps-dmd)

Tips

• Always keep isolated on a transparent background. Without it, prompts that imply a place (e.g. "the Eiffel Tower") get a full scene and the alpha breaks.

• LoRA strength 1.0 (0.8 is fine; at 0.6 subjects turn see-through). With DMD: 8 steps, CFG 1, er_sde/simple. Without DMD: 28 steps, CFG 4.

• Preview images here are flattened onto backgrounds because Civitai doesn't show transparency. Raw RGBA PNGs are on Hugging Face.

How it was made

The RGBA VAE shares FLUX.2's latent space, so it can decode MOSS latents, but plain MOSS latents give garbage alpha. We built a small dataset with MOSS as the image source and alpha "teachers" (Klein Base 4B + Extract-4B on composite/plate pairs, Qwen Image 2.1 for effects), filtered it automatically and trained a rank-32 LoRA at LR 3e-4 with ai-toolkit. Full research notes, training config and the ai-toolkit patch: huggingface.co/paralaif/MOSS-Image-Alpha-LoRA

Licence

Non-commercial research and evaluation use only. Built with Qwen: 8 of the 39 training images were generated with Qwen Image 2.1 (Qwen Research License). Credits: MOSS-Image (OpenMOSS, Apache-2.0), RGBA VAE and Extract-4B (Xavier Jara / trmz, Apache-2.0), FLUX.2 Klein Base 4B (Black Forest Labs, Apache-2.0), ai-toolkit (Ostris). All characters are original designs.