tinypocket commited on
Commit
a453aba
·
verified ·
1 Parent(s): 599dc3b

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +29 -10
README.md CHANGED
@@ -1,5 +1,6 @@
1
  ---
2
- pretty_name: SpeakerKit
 
3
  viewer: false
4
  library_name: whisperkit
5
  tags:
@@ -7,18 +8,36 @@ tags:
7
  - pyannote
8
  - diarization
9
  - speaker-diarization
10
- - whisper
11
- - whisperkit
12
  - coreml
13
- - asr
14
- - quantized
15
- - automatic-speech-recognition
16
  ---
17
 
18
- SpeakerKit Core ML
19
 
20
- Read the [blog](https://www.argmaxinc.com/blog/speakerkit)
 
 
21
 
22
- Try it on [Argmax Playground (TestFlight)](https://testflight.apple.com/join/Q1cywTJw)
 
23
 
24
- Read the [Interspeech 2025 paper](https://www.isca-archive.org/interspeech_2025/durmus25_interspeech.pdf) for architecture and benchmarks.
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
+ license: cc-by-4.0
3
+ pretty_name: akouo — speaker diarization (Core ML)
4
  viewer: false
5
  library_name: whisperkit
6
  tags:
 
8
  - pyannote
9
  - diarization
10
  - speaker-diarization
 
 
11
  - coreml
 
 
 
12
  ---
13
 
14
+ # akouo — speaker diarization (Core ML)
15
 
16
+ Core ML speaker segmentation and embedding models, used by
17
+ [akouo for iOS](https://github.com/tinypocket/akouo-ios) to work out who spoke when — entirely on
18
+ device.
19
 
20
+ This is a **mirror**, hosted so the app does not depend on a third-party repository at runtime.
21
+ The weights are unmodified.
22
 
23
+ ## Provenance
24
+
25
+ - Converted to Core ML by **Argmax**, from
26
+ [argmaxinc/speakerkit-coreml](https://huggingface.co/argmaxinc/speakerkit-coreml).
27
+ - Derived from **pyannote** speaker models.
28
+ - Runs via SpeakerKit, part of [WhisperKit](https://github.com/argmaxinc/WhisperKit) (MIT).
29
+
30
+ ## Licence — attribution required
31
+
32
+ **CC-BY-4.0.**
33
+
34
+ The upstream pyannote models carry mixed terms: `pyannote/segmentation-3.0` is MIT, while
35
+ `pyannote/wespeaker-voxceleb-resnet34-LM` and `pyannote/speaker-diarization-community-1` are
36
+ CC-BY-4.0. Where bundled components differ, the most restrictive governs — so CC-BY-4.0 applies to
37
+ this repository as a whole.
38
+
39
+ **If you redistribute these weights or ship them inside an application, you must credit pyannote
40
+ visibly.** This is a condition of the licence, not a courtesy.
41
+
42
+ Argmax does not declare a licence on the source repository. The licence stated here is a good-faith
43
+ reading of what these weights derive from, not a grant by tinypocket.