Tuning Fork

A reference-tone generator for OVOS - "give me an A" plays a clean 440Hz sine tone, "play 300 hertz" plays an arbitrary frequency. Fully offline, fully deterministic: every tone is generated (equal- temperament, A4 = 440Hz standard concert pitch), nothing recorded, nothing to source or license.

Early 0.0.x release. Named notes always play octave 4 - no octave selection yet. See "Not yet implemented" below.

Usage

"give me an A"
"play a C sharp"
"play 300 hertz"
"giv mig et A"          (Danish)
"spil et H"             (Danish - see "The Danish H/B swap" below)

Each tone rings for 4 seconds (short fade-in/out to avoid clicks) and stops on its own - see "Why no stop intent" below.

The Danish H/B swap - a real trap, not a translation nicety

Danish (like German) uses a different note-naming convention than English for one specific note:

Saying "H" in Danish means the natural B - not a flat, and not "the letter after G" the way English speakers might assume. Saying "B" in Danish means B-flat. This is the actual Danish/German musical convention, the same shape of false-friend trap as the unit- conversion ones in ovos-skill-convert (Danish "ton" vs the English "ton", etc) - getting it backwards would produce a confidently wrong note, not just an awkward translation. See tests/test_note_resolution.py for the regression tests that would catch this being backwards.

The exact spelling of sharp/flat suffixes (ciss vs cis, ess vs es) is lower-confidence - see locale/da-dk/note_aliases.json's _notes key for what's hedged and why.

Why no "stop" intent

Unlike ovos-skill-metronome and ovos-skill-rhythm-box, there's no background thread or stop intent here. play_audio() has no reliable "stop this sound" counterpart in the standard OVOSSkill API (only mycroft.audio.speech.stop, which stops TTS speech, not sounds). Rather than build a stop intent that can't actually stop anything, each tone just has a fixed, short duration and ends on its own - the same way a real tuning fork rings out rather than being interrupted.

Why no "listen and identify the pitch" feature

A natural-feeling extension: strike an instrument, have the skill listen and say whether it's sharp or flat. Deliberately not built - not a scope preference, a real technical blocker: a standard OVOSSkill has no access to raw microphone audio for analysis. The only "listening" primitive (get_response()) goes through the full STT pipeline and returns transcribed text, not a waveform. Real pitch detection would need a dedicated PHAL plugin or audio- transformer tapping the raw audio stream, plus an actual pitch- detection algorithm - a different technical domain (audio analysis) from this skill (audio synthesis), and arguably a different product (an instrument tuner) from this one (a reference-tone generator). Left as a possible future skill, not a feature of this one.

Not yet implemented

• Octave selection. "give me an A" is always A4 (440Hz). Octave- in-speech parsing ("A two" vs "A 2") adds real ambiguity that wasn't worth solving before the basic note lookup worked - use "play {N} hertz" for a specific frequency in the meantime.
• Listen-and-identify pitch - see above, likely a separate future skill rather than an extension of this one.

Install

pip install ovos-skill-tuning-fork

Development

See DEVELOPMENT.md.

Category

Utility

Tags

#music #tuning #reference-tone #instrument