Warlock Studio is a unified platform for upscaling, restoring, denoising, and interpolating frames in videos and images. It is inspired by and based on Djdefrag tools such as QualityScaler and FluidFrames.
📥 Download Installer
This installer was built using PyInstaller and Inno Setup.
By default, it includes DirectML support to ensure maximum compatibility with any graphics card (NVIDIA/AMD/INTEL).
Select your preferred option to download the latest version (Direct Release/SourceForge):
|
|
🆕 New in v6.0 — Process Chaining
- Create multi-step pipelines; order steps to run sequentially per file.
- RIFE interpolation integrates as a chain step for video sources (graceful skip on images).
- Per-step model selection via a combobox fed by auto-discovered ONNX models in
AI-onnx/. - Automatic output routing: intermediate steps use temp folders; the final step writes to your chosen output path.
- Smart extension/codec correction by media type to prevent invalid outputs.
- Memory-safe execution with per-step VRAM tile sizing and cleanup between steps.
🖼️ Interface Capture
🔍 Quality Comparison
✨ Key Features
- AI Upscaling & Restoration – Utilize Real-ESRGAN, BSRGAN, RealESRNet, RealESR_Animex4, and IRCNN models for denoising, super-resolution, and detail recovery.
- Face Restoration (GFPGAN) – Recover facial details from low-resolution or blurry images and video frames.
- Frame Interpolation (RIFE) – Smooth motion or generate slow-motion content with 2×, 4×, or 8× interpolation.
- Process Chaining – Build sequential workflows by chaining steps. Mix upscaling, face restoration, and interpolation; each step’s output becomes the next step’s input automatically. Includes model auto-discovery, per-step GPU/codec settings, and smart validation (e.g., RIFE requires video).
- Advanced Hardware Acceleration – Intelligent provider selection prioritizes CUDA, falls back to DirectML, and finally CPU for maximum compatibility.
- Batch Processing – Process multiple media files simultaneously, saving time and effort.
- Custom Workflows – Fine-grained control over models, resolution, output formats, and quality parameters.
- Open-Source & Extensible – Fully MIT licensed, for contributors and developers.
🖥️ System Requirements
| Component | Minimum Specification | Recommended Specification |
|---|---|---|
| OS | Windows 10 (64-bit) | Windows 11 (64-bit) |
| RAM | 8 GB | 16 GB+ (Required for 4K & High-FPS Video) |
| GPU | DirectX 12 Compatible (DML) / NVIDIA GTX 10-Series | NVIDIA RTX 3060+ / AMD RX 6000+ |
| VRAM | 4 GB | 8 GB - 12 GB+ (For Stable Diffusion/Video Interpolation) |
| Storage | 2 GB available space | SSD (Critical for RIFE & Temp Video Processing) |
| Architecture | x64 | x64 (Native DirectML Support) |
Performance Tip: Given that Warlock Studio leverages DirectML for hardware acceleration, keeping your GPU drivers updated is essential for maximizing processing speed across NVIDIA, AMD, and Intel hardware.
🤝 Contributions
We welcome contributions from the community.
📧 Contact: negroayub97@gmail.com
📜 License & Credits
© 2025 Iván Eduardo Chavez Ayub
Licensed under MIT. Additional terms and attributions are provided in NOTICE.md.
📊 Integrated Technologies & Licenses
| Technology / Model | License | Author / Maintainer | Source |
|---|---|---|---|
| Real-ESRGAN | BSD 3-Clause | Xintao Wang | GitHub |
| • RealESRGANx4 | BSD 3-Clause | Xintao Wang | Same as above |
| • RealESRNetx4 | BSD 3-Clause | Xintao Wang | Same as above |
| • RealESR_Gx4 | BSD 3-Clause | Xintao / Community | Same as above |
| • RealESR_Animex4 | BSD 3-Clause | Community | Same as above |
| BSRGAN | Apache 2.0 | Kai Zhang | GitHub |
| • BSRGANx4 | Apache 2.0 | Kai Zhang | Same as above |
| • BSRGANx2 | Apache 2.0 | Kai Zhang | Same as above |
| IRCNN | BSD / Mixed | Kai Zhang | GitHub |
| • IRCNN_Mx1 | BSD / Mixed | Kai Zhang | Same as above |
| • IRCNN_Lx1 | BSD / Mixed | Kai Zhang | Same as above |
| GFPGAN | Apache 2.0 | TencentARC | GitHub |
| RIFE | MIT | Hzwer / Megvii | GitHub |
| QualityScaler | MIT | Djdefrag | GitHub |
| FluidFrames | MIT | Djdefrag | GitHub |
| DirectML | MIT | Microsoft | GitHub |
| ONNX Runtime | MIT | Microsoft | GitHub |
| CustomTkinter | MIT | Tom Schimansky | GitHub |
| TkinterDnD2 | MIT | pmgagne | GitHub |
| OpenCV (cv2) | Apache 2.0 | OpenCV Team | Official Site |
| NumPy | BSD 3-Clause | NumPy Developers | Official Site |
| Pillow (PIL) | HPND | Python-Pillow Team | GitHub |
| MoviePy | MIT | Zulko | GitHub |
| FFmpeg | LGPL / GPL | FFmpeg Team | Official Site |
| ExifTool | Artistic | Phil Harvey | Official Site |
| Psutil | BSD 3-Clause | Giampaolo Rodola | GitHub |
| WMI | MIT | Tim Golden | GitHub |
| GPUtil | MIT | Anders Krogh | GitHub |
| Requests | Apache 2.0 | Kenneth Reitz | GitHub |
| Packaging | Apache 2.0 | PyPA | GitHub |
| Natsort | MIT | Seth M. Morton | GitHub |
| Python | PSF License | Python Software Foundation | Official Site |
| PyInstaller | GPLv2+ | PyInstaller Team | GitHub |
| Inno Setup | Custom | Jordan Russell | Official Site |