opikdeopay/ads-custom-model

ADS Forge Custom Models & LoRAs Repository

Repositori ini menyimpan seluruh kebutuhan aset bobot model, LoRA, VAE, Text Encoders, dan patch untuk ekosistem ADS Forge (ComfyUI Local & RunPod Cloud Worker).

Struktur direktori di repositori ini didesain 1:1 identik dengan folder instalasi ComfyUI (models/):

├── clip_vision/ ├── diffusion_models/ ├── embeddings/ ├── insightface/ ├── ipadapter/ ├── latent_upscale_models/ ├── loras/ ├── model_patches/ ├── RMBG/ ├── SEEDVR2/ ├── text_encoders/ ├── ultralytics/ ├── upscale_models/ ├── vae/ └── vae_approx/

Ringkasan Kategori Model

1. LoRAs (25 File, ~26.8 GB):
   • MiniMax H3 LoRAs (DMAD 4/8-step, ELM Longlive, FL2VA-Acc, Ref2VA-Acc, Motion/Acceleration Sliders, Rough Sketch, Digital Art)
   • LTX-2.5 & LTX-2.3 LoRAs (Distilled 450, IC-LoRA Ingredients, Refine Details, Colorization, Deblur, Decompression, Pixel Spatial Upscaler, Motion Track, Anime Style)
   • Krea2 Anything2Real Characters
2. Diffusion & Base Models (4 File, ~95.2 GB):
   • MiniMax-H3 Singularity ref2va v2 int8
   • MiniMax-H3 Hybrid Uncensored int8
   • MiniMax-H3 Ref2VA pruned int8
   • Krea2 Turbo bf16
3. Text Encoders (5 File, ~62.8 GB):
   • Gemma 4 12B with proj LTX-2.5 (bf16 & int8)
   • Qwen3-VL 32B MiniMax H3 nvfp4 awq
   • Qwen3-VL 4B bf16
   • Qwen 3 0.6B Base
4. VAE & Latent Upscalers (8 File, ~7.8 GB):
   • LTX-2.5 Video & Audio VAE
   • MiniMax H3 Video & Audio VAE
   • Qwen Image VAE
   • LTX-2.5 & MiniMax H3 Latent Upscalers
5. ControlNet, Patches, Vision & Utility (~15.6 GB):
   • MiniMax H3 Fun ControlNet Union 2.0
   • IP-Adapter FaceID Plus v2 SDXL/SD15 + CLIP Vision ViT-H-14
   • InsightFace Antelopev2
   • RMBG 2.0 BiRefNet & SEEDVR2
   • Ultralytics YOLOv8 (Face & Hand Detailer) & Upscale Models

Cara Download Cepat ke ComfyUI

Gunakan script bawaan ADS Forge atau Python huggingface_hub:

`ash

Download seluruh LoRA

python tools/hf_model_download.py --filter loras

Download seluruh VAE

python tools/hf_model_download.py --filter vae `