gusnion / WaifuStudio

# ✨ WaifuStudio v2.0 — Local Anime & Waifu Studio (Personal Project)
> 🌐 **English** | Español
---
**WaifuStudio** is a personal hobby project I build in my spare time to unify anime image generation, character sheets, **MinimaxH3 (VideoDeltaNet / VDN)** video generation, **Qwen 2.1** image editing, upscaling, and a basic timeline assembler into a single local web interface powered by **ComfyUI** behind the scenes.
> [!WARNING]
> **Disclaimer — No Warranties & Possible Bugs**:
> This is an independent project developed in my free time and provided **"AS IS"**, without guarantees of any kind. It is actively evolving, so **expect bugs, rough edges, or unexpected behavior** (especially on hardware configurations different from mine). Use it at your own risk.
---
## 🎯 Overview of Features
| Module | What It Does |
|---|---|
| ⚡ **Portable Windows Installer** | `INSTALAR.bat` and `ABRIR_APP.bat` set up an isolated Python/ComfyUI environment on Windows and download the configured weights. |
| 🎭 **Character Manager (OC Maker)** | Store character traits, generate Front / Side / Back character sheets, and launch local LoRA training jobs. |
| 🧠 **Prompt Assistant** | 91,357 Danbooru tags indexed in SQLite FTS5 + optional local CPU LLM assistant (`Qwen3.5-9B` GGUF) for tag expansion and vision captioning. |
| 🎞️ **Video H3 (`MinimaxH3 VDN`)** | 8-step video generation using **MinimaxH3 VideoDeltaNet (VDN)** with native audio and camera motion presets (`I2V`, `FLF2V`, `Ref2VA`, `V2V`). |
| 🎬 **Basic Timeline Studio** | Simple web sequencer to stitch generated video clips and static images with an external audio track via FFmpeg, plus shortcuts to continue clips via `I2V` / `FLF2V`. |
| 🔍 **Upscaler & RIFE FPS** | Image and video upscaling (`RealESRGAN`, `AnimeSharp`, `UltraSharp`), optional diffusion pass (`Anima` / `Qwen 2.1`), `RCAS` sharpening, and RIFE frame interpolation. |
---
## 🚀 Quick Start (Windows)
1. **Clone or Download** this repository to your drive (e.g. `E:\IA\WAIFU`).
2. **Run `INSTALAR.bat`**: Sets up the portable Python environment, ComfyUI backend, custom nodes, and downloads the required models.
3. **Run `ABRIR_APP.bat`**: Starts the FastAPI + ComfyUI backend and opens `http://127.0.0.1:8765` in your browser.
> 💡 **Built-in Guide**: Check the **📖 Guide** tab inside the app for a walkthrough of each workspace and its controls.
---
## 🎬 Workspaces Included
### 1. 🎨 Image & OC Maker
- **Dual Model Families**: Supports **Anima 2.9B** models and **SDXL / Illustrious** checkpoints with automatic architecture detection.
- **Prompt Zones**: Splits prompts into *Subject, Clothing, Background, NSFW, and Composition* to keep tags organized.
- **Dual LoraLoader**: Loads LoRAs across both single-UNET (`Anima`) and full Checkpoint (`SDXL/Illustrious`) workflows.
- **Image-to-Tags**: Extracts tags from reference images using WD14 and/or the local CPU LLM.
### 2. 🎞️ Video H3 (MinimaxH3 VDN — VideoDeltaNet)
Generates anime video clips in **8 steps** usin