RestoraX vs SeedVR2

RestoraX is a free, open-source toolkit for restoring old films, home videos, and archival footage, as well as anime, VHS, and newsreels. It describes 25 models for tasks including super-resolution, colorization, face restoration, frame interpolation, scratch and dust removal, deinterlacing, stabilization, SDR-to-HDR conversion, and audio restoration. A visual pipeline builder connects processing steps in a DAG, with typed ports, parallel branches, merge strategies, retry policies, and per-branch progress. RestoraX provides a React web interface, a FastAPI REST API with WebSocket progress, a command-line interface, and a ComfyUI node pack. It is self-hostable, with Docker Compose configurations documented for development and production. Third-party restorers can be added through separate PyPI plugins. Model weights download from HuggingFace Hub on first use. The stated minimums are Python 3.11, 8 GB RAM, 5 GB disk, and FFmpeg; CPU works, while CUDA 12.1 or later with at least 8 GB GPU VRAM is recommended. It is released under the MIT License.

SeedVR2 is an open-source model for restoring video at arbitrary resolutions, with a one-step diffusion-transformer approach that does not add a separate diffusion prior. Its official ComfyUI release can upscale both video and images and provides four nodes to control the pipeline. A standalone command-line interface supports batch and multi-GPU workflows. Inputs may be RGB or RGBA images and video; the documented model choices are 3B and 7B, with FP16, FP8, and quantized GGUF options. Model offloading, VAE tiling, and BlockSwap are available to manage memory during inference. The research describes adaptive window attention, which adjusts window size to output resolution, and adversarial post-training for high-resolution restoration. Standalone installation instructions cover Windows, macOS, and Linux, including Apple Silicon MPS support. Input height and width must be multiples of 32, and the demo crops resized inputs to meet that constraint. SeedVR2 is licensed under Apache 2.0, and released inference code and model weights are linked from the repository. The plan is free.