You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
ArSon Vocal Throw P1: confidence-gated auto BPM from backing-music BUS + beat-phase sync #471
Follow-up to PR #469 ArSon Vocal Throw. This is a user-approved product requirement intentionally not represented as already working in the first native DSP milestone.
Goal
Auto-detect the tempo of music currently playing and align a last-word/phrase vocal delay throw to the actual beat grid, without introducing delay or CPU stalls into the ASIO vocal monitor.
Required architecture
Read an explicitly chosen backing-music/transport BUS or main music source (never infer BPM only from vocal mic bleed). Tap in existing PCM graph with safe off-callback/downsampled analysis, no competing callback / DSP publisher.
Off-ASIO real-time beat onset/periodicity + half/double-time hypothesis handling; track BPM and beat phase with confidence/hysteresis and musical tempo-change stabilization.
One accepted Tempo Master service with source priority: valid song/host transport → stable auto-TRK tempo → TAP/manual fallback. Never flip tempo erratically when intro is soft/no drums or source becomes silent.
Smooth delay retiming and phase-safe cue triggering for 1/4, 1/8, dotted 1/8, half, with precisely defined no-quantize fallback on low confidence.
User controls stay compact: Auto BPM enabled, source BUS selector, BPM readback/confidence and TAP override. Musical cue via MIDI footswitch after native identity & message-thread ACK.
Atomic device/graph epoch versioning to reject stale tempo/trigger; prep ring/FFT or envelope buffer off callback, no dynamic allocation/block/logging on audio thread.
FAT
Real performance at 60/72/95/120/140/180 BPM, live tempo variations, half-time/double-time, quiet intro, backing drum-free ballad, mic bleed and false beats.
ASIO 64/128 frames at 44.1/48/96 kHz, with singer primary monitor unaffected; concurrent WASAPI stream must not back-pressure ASIO.
Phrase-end auto trigger + manual footswitch and silence/active-vocal ducking tested with measured grid offsets & repeat jitter.
Repository housekeeping (2026-11-10): this P1 auto-BPM/beat-phase request duplicates the more detailed engineering and physical acceptance issue #473, which remains OPEN. The M1–M11 software integration was merged through #497/#499, but music-BUS confidence, jitter, singer/ASIO and device-specific end-to-end qualification are not being claimed complete by closing this duplicate. Please follow #473 for the remaining acceptance; original request retained here for history.
Activity
masarray commentedon Nov 10, 2026
Repository housekeeping (2026-11-10): this P1 auto-BPM/beat-phase request duplicates the more detailed engineering and physical acceptance issue #473, which remains OPEN. The M1–M11 software integration was merged through #497/#499, but music-BUS confidence, jitter, singer/ASIO and device-specific end-to-end qualification are not being claimed complete by closing this duplicate. Please follow #473 for the remaining acceptance; original request retained here for history.