An iPhone can recognize the soundtrack playing nearby, recover its current timecode, and start a personal dubbed track at the same moment.
Like glancing at a station clock, the phone learns both which program is playing and where it is now.
ShazamKit matches a fingerprint of the microphone audio to a custom reference signature and returns an updated offset into that recording.
Important boundary: ShazamKit finds identity and time. AVFoundation loads, seeks, buffers, and plays the dubbed audio.
The original soundtrack and every dub need the same zero point and duration map.
Create one ShazamKit signature for the full museum video. Longer signatures provide more opportunities to match.
Export translated tracks against the identical picture timeline. Preserve silence, pauses, and edits.
Map the reference signature to the exhibit ID, version, duration, and URLs for each language.
The match offset anchors both the room video and personal audio to the same source timeline.
Show a listening state and begin playback only after a confident match.
The room player and phone clocks are independent; re-check periodically.
Use headphones. Speaker playback can contaminate microphone capture.
Version the catalog and dubbed assets as one immutable release.
The demo exaggerates drift so the correction is visible. A production app should tune seek and rate thresholds through on-site testing.
AVAudioEngineSHSessionSHCustomCatalogpredictedCurrentMatchOffsetAVPlayer