Skip to content

0.0.12 - 2026-05-21

Choose a tag to compare

View all tags
@github-actions github-actions released this 21 May 14:47
v0.0.12
63a3792

Release Notes

OSTT 0.0.12 is a substantial release focused on making local, offline transcription a first-class workflow while keeping the existing cloud-provider flexibility intact. You can now download and manage local Whisper-compatible models from the terminal, run transcription fully offline, and use GPU-accelerated builds where supported.

Added

  • Local transcription with whisper.cpp — OSTT can now transcribe audio without sending it to a cloud provider. Local transcription works with curated models from the OSTT model registry as well as custom Hugging Face or direct .gguf / ggml-*.bin model URLs.
  • GPU-accelerated local builds — Metal is enabled on all macOS builds, including Apple Silicon and Intel Macs. Linux x86_64 users can choose CUDA builds for NVIDIA GPUs or Vulkan builds for AMD/Intel GPUs, with CPU builds still available for maximum compatibility.
  • Automatic GPU-aware installer — The installer detects platform, architecture, and available GPU runtime support, then selects the best matching artifact. It prefers CUDA for usable NVIDIA setups, Vulkan for AMD/Intel Linux GPUs, Metal on macOS, and CPU builds otherwise.
  • Local model management UI — ostt model now supports browsing available local models, downloading them, activating a model, deleting installed models, and adding custom model sources. Cloud models also have a clearer info screen with model ID, description, language coverage, and active status.
  • Persistent local model daemon — ostt daemon can keep a local model loaded between transcriptions, avoiding repeated model startup cost. It is managed with ostt daemon start|stop|restart|status|install|uninstall and installs as a launchd service on macOS or a systemd user service on Linux.
  • Daemon-aware transcription path — Local transcription now uses the daemon automatically when it is running with the selected model, and falls back to in-process transcription when needed. Logs make it clear whether a transcription used the daemon or in-process mode.
  • Mistral provider — Added Mistral Voxtral support with Voxtral Mini Transcribe and Voxtral Mini 2602 models. Mistral supports optional language hints through [providers.mistral].language.

Changed

  • Improved model picker experience — Local and cloud model screens now use consistent terminal styling, clearer dialogs, and better toasts.
  • Better provider/model metadata guardrails — Tests now require supported cloud models to define complete metadata, reducing the chance of incomplete model entries shipping in a release.
  • Release artifacts now include GPU variants — In addition to the standard CPU archives and native packages, the release includes CUDA and Vulkan Linux x86_64 archives plus .deb and .rpm packages.

Install ostt 0.0.12

Install prebuilt binaries via shell script

curl --proto '=https' --tlsv1.2 -LsSf https://github.com/kristoferlund/ostt/releases/download/v0.0.12/ostt-installer.sh | sh

Install prebuilt binaries via Homebrew

brew install kristoferlund/ostt/ostt

Download ostt 0.0.12

File Platform Checksum
ostt-aarch64-apple-darwin.tar.gz Apple Silicon macOS checksum
ostt-x86_64-apple-darwin.tar.gz Intel macOS checksum
ostt-aarch64-unknown-linux-gnu.tar.gz ARM64 Linux checksum
ostt-x86_64-unknown-linux-gnu.tar.gz x64 Linux checksum
Assets 40
Loading