1.0.0 RC 1000

Release candidate of Voxprint AI Movie Dubber: 1.0.0 RC · build 1000 "Chazak" ("be strong", Joshua 1:9: steady and resolved before the first test on real hardware).

Requirements: Windows 11 24H2 or newer (64-bit), an NVIDIA GeForce RTX 40-series graphics card or newer and an NVIDIA driver from the 600 branch or newer. Setup and the program stop with a clear message on older hardware.

Two installers - Full installer (offline, recommended): download VoxprintMovieDubber-Full-Setup.exe and every VoxprintMovieDubber-Full-Setup-N.bin part into one folder, then start the .exe. It carries Python 3.14, PyTorch 2.11.0+cu130 and every other component, so setup needs no internet. The AI models (about 8 GB) are downloaded by the optional setup task or on first use. - Online installer: the small VoxprintMovieDubber-Setup.exe downloads Python, PyTorch and the other components during setup.

Both install the same shared runtime, so switching between them later downloads nothing twice. Check every file against the SHA-256 table below (also in the .sha256 files).

What's new since build 999 "Hineni" - Python 3.14 and PyTorch 2.11.0 with CUDA 13 (cu130); RTX 40-series or newer with a 600+ driver is required, and there is no processor-only mode. - Thermal rule of the Voxprint suite: full speed for the first 2.5 hours of a dub, then a pause between batches only while the 5-minute median GPU temperature is 83 °C or more, or the card keeps throttling for more than 60 seconds; work resumes at 75 °C. The 2.5 hours count from the start of the whole dub. Fully automatic, no setting. - VRAM-aware model tiers: the setup and the program detect the card's video memory and pick a 16 GB or a 24 GB+ tier by themselves (lines per speech batch, models offered). Models the card cannot run are greyed out; Settings has an optional advanced override that can only pick a tier the card supports. On cards too small for the 1.7B speech model, setup fetches the 0.6B model instead. - .vxdub project files: Save, Save As and Open, with their own icon; double-click a project to continue where you stopped. - Model inventory: every model pinned to an exact revision with checksums, newer Opus-MT tc-big translation models, licences and CC-BY attributions recorded. - New install and usage instructions in the README and a step-by-step User guide. - Build: GitHub Actions builds, installs, tests and uninstalls both installers before a release; actions/checkout v7 and actions/setup-python v7.