VRGDG SeedVR2 TensorRT Studio Released — A Video Restoration Studio

VRGDG SeedVR2 TensorRT Studio, a local open-source tool for video restoration and upscaling on NVIDIA RTX, has been released on GitHub in the vrgamegirl19 repository.

Compiled by Sergey KostenchukPublished 2026-09-01Updated 2026-09-18

2026-09-01 Research NVIDIA

🤖 VRGDG SeedVR2 TensorRT Studio Released — A Video Restoration Studio

The open-source tool by vrgamegirl19 (GitHub, 81 stars, August 26, 2026) restores and upscales video using the SeedVR2 pipeline with TensorRT acceleration on NVIDIA RTX. It features a preview before the full render, a slider comparison, resuming rendering from a chunk, and post-processing without re-rendering. A stated example: an 8-second 360p→2K clip in ~8 minutes on an RTX 5090.

🌍 SeedVR2 was a research CLI pipeline — it now has a GUI workflow with checkpoints and a TensorRT decoder: a typical path for turning research models into video production tools.

👤 With an RTX card and Windows, you can locally and for free restore old or "washed-out" video: installation via two .bat files, preview, comparison with the original, and sharpness fine-tuning without re-rendering. FP8 saves VRAM, while FP16 delivers maximum quality.

Source 1: https://github.com/vrgamegirl19/VRGDG-SeedVR2-TensorRT-Studio

Sources