Skip to content

LRC Bulk Finder v2.9.4 — English / Ελληνικά (Public Beta)

Choose a tag to compare

@github-actions github-actions released this 23 Oct 07:58
d6c4279

v2.9.4 — One-click Greek G2G AI setup for Windows

Why the checkbox was disabled

The standalone app does not bundle AUEB's 1.2 GB Greeklish-to-Greek model, Python, PyTorch, or other heavy dependencies. In v2.9.3 the GUI protected users from selecting an AI mode that would inevitably fail. v2.9.4 adds a dedicated, optional installer and runtime companion.

New behavior

  • Click Install / verify Greek G2G AI (one click) inside Saved searches → Downloads & local setup.
  • Read/accept the download confirmation. No command-prompt navigation, manual folders or hand editing of PATH are required.
  • When necessary, the installer searches for Python 3.10–3.13 and can use Windows winget to install official Python 3.11. Windows/winget may request separate permissions.
  • Creates an isolated environment under Local AppData/LRCBulkFinder/G2G, installs CPU-oriented PyTorch and the upstream gr-nlp-toolkit==0.3.0, downloads the full G2G model and runs an actual conversion test.
  • A verified marker is written only after the self-test succeeds. Then the AI checkbox and Greek Local AI FULL preset become available in the standalone GUI. Failure or interruption keeps the option disabled; dictionary-only Greek search still works.
  • A persistent local Python companion handles AI conversions in the search worker without loading the large model repeatedly; inference is CPU-based so modest 4 GB GPUs are not required. Once installed, search-time model loading uses offline HF settings.
  • The progress window shows installer output; the app refuses to begin a music search while installing.

Safety

No installation begins without clicking the button and explicitly approving the network/download notice. No audio files, existing LRCs or ID3 metadata are changed by installation. The AI generates text SEARCH SUGGESTIONS and does not verify song/lyrics matches. A separate user environment avoids altering system-wide Python. Upstream downloads and optional winget/Python setup require internet and disk space and are subject to their own licenses and platforms.

Limitations and QA

  • The full model is ~1.2 GB plus potentially multiple GB of dependencies and may use substantial CPU/RAM. It is not included in the ZIP.
  • Windows PowerShell/winget availability and antivirus settings vary; the GUI reports real failures rather than claiming setup success.
  • GitHub offline tests mock installers/models and verify readiness/fallback. Frozen-EXE tests validate entrypoints, worker status and packaging. The first real model download/inference still requires acceptance testing on a user's Windows computer.
  • This is a beta. Existing v2.9.3 Windows release remains intact, and all previous Greek dictionaries, 15 search presets, UTF-8 fixes, STOP protection and no-overwrite logic are retained.