Upgrade background removal to BiRefNet Lucida

Add Muse v3 Krea 2 checkpoint

Fix ComfyUI torchaudio imports

Make Z Image asset downloads reliable

Add LumiSemiReal Krea2 LoRA

Use Krea2 TextGenerate prompt enhancement

Add RMBG model dependency

Add ComfyUI RMBG background removal

Add Krea2 yunyun LoRA

Use Krea2 text encode node

Auto translate Chinese prompts for Anima

Force SpiceNiji as Anima LoRA

Move SpiceNiji back to Anima slot

Resume large direct downloads

Retry direct model downloads safely

Keep Krea2 LoRA slot separate

Route SpiceNiji through Krea2 LoRA slot

Add Anima LoRA from HF bucket

Use HF bucket for Anima checkpoint

Fix Krea CivitAI download URL

Release inactive models before switching

Use user CivitAI Anima and Krea models

Add optional Krea CivitAI LoRA

Use official Anima split files

Remove Lucid and fix Anima Krea loading

Align Krea sampler with official workflow

Default Krea prompt enhance off

Fix Krea 2 VAE sampler and prompt enhance

Skip unused ERNIE LoRA downloads

Use Huihui text encoder for Krea 2

Add Krea 2 model backend

Add Anima model backend

feat: dynamic LoRA registry from lora_meta.json (drop file in dataset = auto-show)

refactor: pull all 6 LoRAs from jing96963/loras dataset (single source of truth)

fix: lower @spaces.GPU duration from 180 to 120 (Pro per-call cap)

Upload app.py with huggingface_hub

Upload app.py with huggingface_hub

Upload app.py with huggingface_hub

Upload app.py with huggingface_hub

Upload app.py with huggingface_hub

Upload app.py with huggingface_hub

Upload app.py with huggingface_hub

Upload app.py with huggingface_hub

Upload app.py with huggingface_hub

Upload requirements.txt with huggingface_hub

Upload app.py with huggingface_hub

Upload app.py with huggingface_hub

Upload app.py with huggingface_hub

Upload app.py with huggingface_hub

Upload requirements.txt with huggingface_hub