InfernoBench v1.2.0 — SeedVR2 Restoration

SeedVR2 enters the bench

• Upload and restore video locally with ByteDance SeedVR2-3B.
• One-command installer for the verified 3B FP8 consumer-GPU path.
• Full CPU BlockSwap and tiled VAE defaults for 8–16 GB NVIDIA GPUs.
• Preview (480p) and HD (720p) restoration modes in Inferno Studio.
• Queued progress, browser playback, MP4 download, and input-audio preservation.
• Pinned runtime/model revisions with SHA-256 verification.
• Verified on an RTX 5060 Ti 16 GB with real 5-frame and 33-frame video tests.

Install SeedVR2:

curl -fsSL https://raw.githubusercontent.com/GPUCLOUDXYZ/inferno-bench/main/bootstrap-seedvr2.sh | sh

See the README for hardware limits, test results, and the beginner walkthrough.

InfernoBench v1.1.1 — the forge has a name

Sulphur has grown into InfernoBench

InfernoBench is the new umbrella for open-source model installers, reproducible local inference benchmarks, and browser-based test labs. Sulphur Studio remains the video UI, with Wan2.1 and Sulphur-2 as its first engines.

Install Wan2.1

curl -fsSL https://raw.githubusercontent.com/GPUCLOUDXYZ/inferno-bench/main/bootstrap-wan.sh | sh

Install Sulphur-2 with audio

curl -fsSL https://raw.githubusercontent.com/GPUCLOUDXYZ/inferno-bench/main/bootstrap.sh | sh

This patch updates canonical repository URLs, badges, bootstrap download sources, AI-readable metadata, and versioned install directories. Existing GitHub links redirect to the new repository.

Sulphur v1.1.0 — Wan2.1 joins the studio

Wan2.1 in the shared local video UI

Install the new Wan2.1 edition with one command:

curl -fsSL https://raw.githubusercontent.com/GPUCLOUDXYZ/sulphur/main/bootstrap-wan.sh | sh

What is new

• Native Wan-AI/Wan2.1-T2V-1.3B-Diffusers engine pinned to commit 0fad780a534b6463e45facd96134c9f345acfa5b
• Wan2.1 / Sulphur-2 engine selector in the same responsive LAN UI
• Queued local Diffusers worker with CPU offload and live sampling progress
• Draft: 33 frames, 512×320, 16 FPS; Standard: 81 frames, 832×480, 16 FPS
• Independent UI runtime and optional engine services
• Updated health checks, sulphurctl, screenshot, noob guide, and AI-discovery metadata

Verified on the release host

RTX 5060 Ti 16 GB, PyTorch 2.13.0 + CUDA 13.0, Diffusers 0.39.0. A real Draft request through the browser API produced a 2.06-second H.264 MP4. Cold request: about 12 minutes from external SSD; warm Draft: about 39 seconds.

Wan2.1 output is silent. Select Sulphur-2 when generated audio is required.

Sulphur v1.0.2 — one-line installer

Install Sulphur with one command

curl -fsSL https://raw.githubusercontent.com/GPUCLOUDXYZ/sulphur/main/bootstrap.sh | sh

The new streamed bootstrap downloads this pinned release, performs the read-only GPU/RAM/disk preflight, requests administrator access, and launches the complete Sulphur-2 + LTX-2.3 + ComfyUI installation. Interrupted model downloads remain resumable.

Sulphur v1.0.1 — One-command local AI video setup

A clearer, beginner-first release for local Sulphur-2 text-to-video generation. Adds a prominent one-command installer, real UI preview, improved hardware and testing guidance, GitHub discovery metadata, and an AI-readable llms.txt project summary.