Euphonia v0.2.0 — zero-setup insights
Every take now gets a written insight instantly, with no API key and no setup at all.
What changed
- Insights are generated locally by a built-in rule-based writer that reuses the app's own metric zones (pitch, resonance, steadiness, vocal weight, phrasing) to pick what's working and the single clearest thing to try next — no network call, no account, nothing to configure.
- A free Gemini API key is now a purely optional upgrade, not a setup step. Add one anytime from the ⚙️ Settings screen, and any take showing the built-in insight will offer a "✨ upgrade to an AI-written insight" button for a richer, more personalized version.
- The first-launch API key prompt is gone — the app is fully usable the moment it's installed.
We also prototyped running a small local LLM (Qwen3, via HuggingFace) as a fully-offline replacement for Gemini, including a zone-aware prompt that explicitly tells the model which direction each metric is good/bad in. It wasn't reliable enough to trust for coaching advice, so we went with the deterministic built-in writer above instead — same zero-setup benefit, no risk of misreading your numbers.
Getting started
- Download
Euphonia Setup 0.2.0.exebelow. - Run it — Windows SmartScreen will warn it's from an unrecognized publisher (it's unsigned; not a sign anything's broken). Click More info → Run anyway.
- Record a take. That's it — you'll get a written insight right away, no key required.
- (Optional) Add a free Gemini API key from the ⚙️ Settings screen any time for richer, AI-written insights. See docs/gemini-api-key.md for how to get one.
Requirements
- Windows only for this release (see README for build-from-source options on other platforms)
Licensing note
This installer bundles GPLv3 components (Praat/parselmouth, ffmpeg). The source in this repo is MIT/public domain, but the packaged installer as a combined work is GPLv3 — see the README's License section and the bundled THIRD-PARTY-LICENSES.md for details.
Full documentation: README.md