feat(speakers): human-in-the-loop cross-recording speaker identification - #472

Merged

Optic00 merged 22 commits into

stenolabs:feat/speaker-diarizationfrom

valentinweyer:speaker-identity

Aug 4, 2026

Commits on Jul 30, 2026

feat(transcribe): per-channel acoustic speaker diarization
valentinweyer
committed
e310a51
feat(transcribe): acoustic diarization for mono audio
valentinweyer
committed
f9fcfae
perf(diarize): ~6.65x faster via Sortformer's offline chunk preset
valentinweyer
committed
11dc570
feat(diarize): robust sidecar parsing, dominance gate, live progress
valentinweyer
committed
7fcde29
feat(processing): live per-stage progress for diarization
valentinweyer
committed
9ea3dcb
fix(e2e): seed a real recording session in processing-stages.t1
valentinweyer
committed
f31f2c9
feat(speakers): human-in-the-loop speaker identification
valentinweyer
committed
f1e1b71
feat(diarize): restore voiceprint embedding extraction to the sidecar
valentinweyer
committed
0741c18
fix(backend-cli): PYTHONUNBUFFERED:1 on every spawned backend process
valentinweyer
committed
842f9f5
feat(processing): restore embedding sub-progress + transcribe progress
valentinweyer
committed
e0621f2
feat(main): throttle PROGRESS:transcribe/embedding lines in the disk log
valentinweyer
committed
e0bb5bf
feat(transcriber): restore identity-side diarization additions
valentinweyer
committed
56ead44
wip: voiceprint/diarization work carried over from local/mic-plus-voiceprint
valentinweyer
committed
1d87b6a
fix(speakers): channel-scope prototype matching, correction path, relative hard-negatives
valentinweyer
committed
9d89027
feat(speakers): enroll-self-from-person CLI
valentinweyer
committed
ea6ddad
fix(e2e): seed a real recording session in processing-stages.t1
valentinweyer
committed
166f83b

Commits on Jul 31, 2026

fix(speakers): verified deletion + a setting to disable identity matching
valentinweyer
committed
7e20f0f

Commits on Aug 1, 2026

Commits on Aug 3, 2026

Merge remote-tracking branch 'upstream/feat/speaker-diarization' into speaker-identity
valentinweyer
committed
b54b429

Commits on Aug 4, 2026

Merge branch 'feat/speaker-diarization' into speaker-identity
Optic00
committed
fab3bba

Navigation Menu

Uh oh!

Uh oh!

feat(speakers): human-in-the-loop cross-recording speaker identification - #472

feat(speakers): human-in-the-loop cross-recording speaker identification#472
Optic00 merged 22 commits into
stenolabs:feat/speaker-diarizationfrom
valentinweyer:speaker-identity

Commits on Jul 30, 2026

feat(transcribe): per-channel acoustic speaker diarization

feat(transcribe): acoustic diarization for mono audio

perf(diarize): ~6.65x faster via Sortformer's offline chunk preset

feat(diarize): robust sidecar parsing, dominance gate, live progress

feat(processing): live per-stage progress for diarization

fix(e2e): seed a real recording session in processing-stages.t1

feat(speakers): human-in-the-loop speaker identification

feat(diarize): restore voiceprint embedding extraction to the sidecar

fix(backend-cli): PYTHONUNBUFFERED:1 on every spawned backend process

feat(processing): restore embedding sub-progress + transcribe progress

feat(main): throttle PROGRESS:transcribe/embedding lines in the disk log

feat(transcriber): restore identity-side diarization additions

wip: voiceprint/diarization work carried over from local/mic-plus-voiceprint

fix(speakers): channel-scope prototype matching, correction path, relative hard-negatives

feat(speakers): enroll-self-from-person CLI

fix(e2e): seed a real recording session in processing-stages.t1

Commits on Jul 31, 2026

fix(speakers): verified deletion + a setting to disable identity matching

Commits on Aug 1, 2026

fix(e2e): make isSayAvailable() check real synthesis, fix impossible invariant

docs: update speaker-label docs for per-channel acoustic diarization

fix(backend-cli): attach stdout/stderr to non-zero-exit rejections

Commits on Aug 3, 2026

Merge remote-tracking branch 'upstream/feat/speaker-diarization' into speaker-identity

Commits on Aug 4, 2026

Merge branch 'feat/speaker-diarization' into speaker-identity