aotrih commited on
Commit
c920217
·
verified ·
1 Parent(s): 082165c

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +6 -123
README.md CHANGED
@@ -1,6 +1,6 @@
1
 
2
  ---
3
- pretty_name: "WhisperKit ASR Evaluation Results"
4
  viewer: false
5
  library_name: whisperkit
6
  tags:
@@ -10,128 +10,11 @@ tags:
10
  - asr
11
  - quantized
12
  ---
13
- # WhisperKit Transcription Quality
14
 
 
 
15
 
16
-
17
- ## Dataset: `librispeech`
18
- Short-form Audio (<30s/clip) - 5 hours of English audiobook clips
19
-
20
- | | WER (↓) | QoI (↑) | File Size (MB) | Code Commit |
21
- |:------------------------------------------------------------------------------------------------------------------------------|:--------------------------------------------------------------------------------------------------------------------------------------|----------:|-----------------:|:---------------------------------------------------------------|
22
- | large-v2 (WhisperOpenAIAPI) | [2.35](https://hf.co/datasets/argmaxinc/whisperkit-evals/tree/main/WhisperOpenAIAPI/openai_whisper-large-v2/librispeech) | 100 | 3100 | N/A |
23
- | [large-v2](https://hf.co/argmaxinc/whisperkit-coreml/tree/main/openai_whisper-large-v2) | [2.77](https://hf.co/datasets/argmaxinc/whisperkit-evals/tree/main/WhisperKit/openai_whisper-large-v2/librispeech) | 96.6 | 3100 | [Link](https://github.com/argmaxinc/WhisperKit/commit/2846fd9) |
24
- | [large-v2_949MB](https://hf.co/argmaxinc/whisperkit-coreml/tree/main/openai_whisper-large-v2_949MB) | [2.4](https://hf.co/datasets/argmaxinc/whisperkit-evals/tree/main/WhisperKit/openai_whisper-large-v2_949MB/librispeech) | 94.6 | 949 | [Link](https://github.com/argmaxinc/WhisperKit/commit/eca4a2e) |
25
- | [large-v2_turbo](https://hf.co/argmaxinc/whisperkit-coreml/tree/main/openai_whisper-large-v2_turbo) | [2.76](https://hf.co/datasets/argmaxinc/whisperkit-evals/tree/main/WhisperKit/openai_whisper-large-v2_turbo/librispeech) | 96.6 | 3100 | [Link](https://github.com/argmaxinc/WhisperKit/commit/2846fd9) |
26
- | [large-v2_turbo_955MB](https://hf.co/argmaxinc/whisperkit-coreml/tree/main/openai_whisper-large-v2_turbo_955MB) | [2.41](https://hf.co/datasets/argmaxinc/whisperkit-evals/tree/main/WhisperKit/openai_whisper-large-v2_turbo_955MB/librispeech) | 94.6 | 955 | [Link](https://github.com/argmaxinc/WhisperKit/commit/cf75348) |
27
- | [large-v3](https://hf.co/argmaxinc/whisperkit-coreml/tree/main/openai_whisper-large-v3) | [2.04](https://hf.co/datasets/argmaxinc/whisperkit-evals/tree/main/WhisperKit/openai_whisper-large-v3/librispeech) | 95.2 | 3100 | [Link](https://github.com/argmaxinc/WhisperKit/commit/2846fd9) |
28
- | [large-v3_turbo](https://hf.co/argmaxinc/whisperkit-coreml/tree/main/openai_whisper-large-v3_turbo) | [2.03](https://hf.co/datasets/argmaxinc/whisperkit-evals/tree/main/WhisperKit/openai_whisper-large-v3_turbo/librispeech) | 95.4 | 3100 | [Link](https://github.com/argmaxinc/WhisperKit/commit/2846fd9) |
29
- | [large-v3_turbo_954MB](https://hf.co/argmaxinc/whisperkit-coreml/tree/main/openai_whisper-large-v3_turbo_954MB) | [2.47](https://hf.co/datasets/argmaxinc/whisperkit-evals/tree/main/WhisperKit/openai_whisper-large-v3_turbo_954MB/librispeech) | 93.9 | 954 | [Link](https://github.com/argmaxinc/WhisperKit/commit/cf75348) |
30
- | [distil-large-v3](https://hf.co/argmaxinc/whisperkit-coreml/tree/main/distil-whisper_distil-large-v3) | [2.47](https://hf.co/datasets/argmaxinc/whisperkit-evals/tree/main/WhisperKit/distil-whisper_distil-large-v3/librispeech) | 89.7 | 1510 | [Link](https://github.com/argmaxinc/WhisperKit/commit/cf75348) |
31
- | [distil-large-v3_594MB](https://hf.co/argmaxinc/whisperkit-coreml/tree/main/distil-whisper_distil-large-v3_594MB) | [2.96](https://hf.co/datasets/argmaxinc/whisperkit-evals/tree/main/WhisperKit/distil-whisper_distil-large-v3_594MB/librispeech) | 85.4 | 594 | [Link](https://github.com/argmaxinc/WhisperKit/commit/508240f) |
32
- | [distil-large-v3_turbo](https://hf.co/argmaxinc/whisperkit-coreml/tree/main/distil-whisper_distil-large-v3_turbo) | [2.47](https://hf.co/datasets/argmaxinc/whisperkit-evals/tree/main/WhisperKit/distil-whisper_distil-large-v3_turbo/librispeech) | 89.7 | 1510 | [Link](https://github.com/argmaxinc/WhisperKit/commit/508240f) |
33
- | [distil-large-v3_turbo_600MB](https://hf.co/argmaxinc/whisperkit-coreml/tree/main/distil-whisper_distil-large-v3_turbo_600MB) | [2.78](https://hf.co/datasets/argmaxinc/whisperkit-evals/tree/main/WhisperKit/distil-whisper_distil-large-v3_turbo_600MB/librispeech) | 86.2 | 600 | [Link](https://github.com/argmaxinc/WhisperKit/commit/ae1cf96) |
34
- | [small.en](https://hf.co/argmaxinc/whisperkit-coreml/tree/main/openai_whisper-small.en) | [3.12](https://hf.co/datasets/argmaxinc/whisperkit-evals/tree/main/WhisperKit/openai_whisper-small.en/librispeech) | 85.8 | 483 | [Link](https://github.com/argmaxinc/WhisperKit/commit/228630c) |
35
- | [small](https://hf.co/argmaxinc/whisperkit-coreml/tree/main/openai_whisper-small) | [3.45](https://hf.co/datasets/argmaxinc/whisperkit-evals/tree/main/WhisperKit/openai_whisper-small/librispeech) | 83 | 483 | [Link](https://github.com/argmaxinc/WhisperKit/commit/228630c) |
36
- | [base.en](https://hf.co/argmaxinc/whisperkit-coreml/tree/main/openai_whisper-base.en) | [3.98](https://hf.co/datasets/argmaxinc/whisperkit-evals/tree/main/WhisperKit/openai_whisper-base.en/librispeech) | 75.3 | 145 | [Link](https://github.com/argmaxinc/WhisperKit/commit/228630c) |
37
- | [base](https://hf.co/argmaxinc/whisperkit-coreml/tree/main/openai_whisper-base) | [4.97](https://hf.co/datasets/argmaxinc/whisperkit-evals/tree/main/WhisperKit/openai_whisper-base/librispeech) | 67.2 | 145 | [Link](https://github.com/argmaxinc/WhisperKit/commit/228630c) |
38
- | [tiny.en](https://hf.co/argmaxinc/whisperkit-coreml/tree/main/openai_whisper-tiny.en) | [5.61](https://hf.co/datasets/argmaxinc/whisperkit-evals/tree/main/WhisperKit/openai_whisper-tiny.en/librispeech) | 63.9 | 66 | [Link](https://github.com/argmaxinc/WhisperKit/commit/228630c) |
39
- | [tiny](https://hf.co/argmaxinc/whisperkit-coreml/tree/main/openai_whisper-tiny) | [7.47](https://hf.co/datasets/argmaxinc/whisperkit-evals/tree/main/WhisperKit/openai_whisper-tiny/librispeech) | 52.5 | 66 | [Link](https://github.com/argmaxinc/WhisperKit/commit/228630c) |
40
- | [large-v3-v20240930](https://hf.co/argmaxinc/whisperkit-coreml/tree/main/openai_whisper-large-v3-v20240930) | [1.94](https://hf.co/datasets/argmaxinc/whisperkit-evals/tree/main/WhisperKit/openai_whisper-large-v3-v20240930/librispeech) | 93.9 | 1640 | [Link](https://github.com/argmaxinc/WhisperKit/commit/c2f1b57) |
41
- | [large-v3-v20240930_626MB](https://hf.co/argmaxinc/whisperkit-coreml/tree/main/openai_whisper-large-v3-v20240930_626MB) | [1.95](https://hf.co/datasets/argmaxinc/whisperkit-evals/tree/main/WhisperKit/openai_whisper-large-v3-v20240930_626MB/librispeech) | 93.8 | 626 | [Link](https://github.com/argmaxinc/WhisperKit/commit/3cd3ef1) |
42
-
43
- ## Dataset: `earnings22`
44
- Long-Form Audio (>1hr/clip) - 120 hours of earnings call recordings in English with various accents
45
-
46
- | | WER (↓) | QoI (↑) | File Size (MB) | Code Commit |
47
- |:------------------------------------------------------------------------------------------------------|:--------------------------------------------------------------------------------------------------------------------------|----------:|-----------------:|:---------------------------------------------------------------|
48
- | large-v2 (WhisperOpenAIAPI) | [16.27](https://hf.co/datasets/argmaxinc/whisperkit-evals/tree/main/WhisperOpenAIAPI/openai_whisper-large-v2/earnings22) | 100 | 3100 | N/A |
49
- | [large-v3](https://hf.co/argmaxinc/whisperkit-coreml/tree/main/openai_whisper-large-v3) | [15.17](https://hf.co/datasets/argmaxinc/whisperkit-evals/tree/main/WhisperKit/openai_whisper-large-v3/earnings22) | 58.5 | 3100 | [Link](https://github.com/argmaxinc/WhisperKit/commit/2846fd9) |
50
- | [distil-large-v3](https://hf.co/argmaxinc/whisperkit-coreml/tree/main/distil-whisper_distil-large-v3) | [15.28](https://hf.co/datasets/argmaxinc/whisperkit-evals/tree/main/WhisperKit/distil-whisper_distil-large-v3/earnings22) | 46.3 | 1510 | [Link](https://github.com/argmaxinc/WhisperKit/commit/508240f) |
51
- | [base.en](https://hf.co/argmaxinc/whisperkit-coreml/tree/main/openai_whisper-base.en) | [23.49](https://hf.co/datasets/argmaxinc/whisperkit-evals/tree/main/WhisperKit/openai_whisper-base.en/earnings22) | 6.5 | 145 | [Link](https://github.com/argmaxinc/WhisperKit/commit/dda6571) |
52
- | [tiny.en](https://hf.co/argmaxinc/whisperkit-coreml/tree/main/openai_whisper-tiny.en) | [28.64](https://hf.co/datasets/argmaxinc/whisperkit-evals/tree/main/WhisperKit/openai_whisper-tiny.en/earnings22) | 5.7 | 66 | [Link](https://github.com/argmaxinc/WhisperKit/commit/dda6571) |
53
-
54
- ## Dataset: `common_voice_17_0-argmax_subset-400`
55
- Short-form Audio (<30s/clip) - Max 400 samples per language from Common Voice 17.0 Test Set
56
-
57
- | | es | ro | th | nl | id | sv | de | pl | fi | it | cs | en | vi | el | hu | ru | gl | fr | pt | da | File Size (MB) | Code Commit |
58
- |:--------------------------------------------------------------------------------------------------------------------------------------------------|:----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|-----------------:|:---------------------------------------------------------------|
59
- | [large-v3](https://hf.co/argmaxinc/whisperkit-coreml/tree/main/openai_whisper-large-v3) | [4.93](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3/common_voice_17_0-argmax_subset-400/forced/es) | [5.39](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3/common_voice_17_0-argmax_subset-400/forced/ro) | [6.11](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3/common_voice_17_0-argmax_subset-400/forced/th) | [7.03](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3/common_voice_17_0-argmax_subset-400/forced/nl) | [9.47](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3/common_voice_17_0-argmax_subset-400/forced/id) | [9.81](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3/common_voice_17_0-argmax_subset-400/forced/sv) | [9.89](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3/common_voice_17_0-argmax_subset-400/forced/de) | [10.13](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3/common_voice_17_0-argmax_subset-400/forced/pl) | [10.32](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3/common_voice_17_0-argmax_subset-400/forced/fi) | [11.11](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3/common_voice_17_0-argmax_subset-400/forced/it) | [12.04](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3/common_voice_17_0-argmax_subset-400/forced/cs) | [12.21](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3/common_voice_17_0-argmax_subset-400/forced/en) | [12.32](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3/common_voice_17_0-argmax_subset-400/forced/vi) | [12.35](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3/common_voice_17_0-argmax_subset-400/forced/el) | [12.44](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3/common_voice_17_0-argmax_subset-400/forced/hu) | [13.0](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3/common_voice_17_0-argmax_subset-400/forced/ru) | [13.06](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3/common_voice_17_0-argmax_subset-400/forced/gl) | [13.67](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3/common_voice_17_0-argmax_subset-400/forced/fr) | [13.75](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3/common_voice_17_0-argmax_subset-400/forced/pt) | [13.89](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3/common_voice_17_0-argmax_subset-400/forced/da) | 3100 | [Link](https://github.com/argmaxinc/WhisperKit/commit/e3e21d4) |
60
- | [large-v2](https://hf.co/argmaxinc/whisperkit-coreml/tree/main/openai_whisper-large-v2) | [6.93](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v2/common_voice_17_0-argmax_subset-400/forced/es) | [7.86](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v2/common_voice_17_0-argmax_subset-400/forced/ro) | [8.76](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v2/common_voice_17_0-argmax_subset-400/forced/th) | [8.93](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v2/common_voice_17_0-argmax_subset-400/forced/nl) | [12.2](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v2/common_voice_17_0-argmax_subset-400/forced/id) | [12.16](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v2/common_voice_17_0-argmax_subset-400/forced/sv) | [11.7](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v2/common_voice_17_0-argmax_subset-400/forced/de) | [12.51](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v2/common_voice_17_0-argmax_subset-400/forced/pl) | [13.13](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v2/common_voice_17_0-argmax_subset-400/forced/fi) | [14.34](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v2/common_voice_17_0-argmax_subset-400/forced/it) | [17.14](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v2/common_voice_17_0-argmax_subset-400/forced/cs) | [12.7](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v2/common_voice_17_0-argmax_subset-400/forced/en) | [17.69](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v2/common_voice_17_0-argmax_subset-400/forced/vi) | [15.04](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v2/common_voice_17_0-argmax_subset-400/forced/el) | [16.72](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v2/common_voice_17_0-argmax_subset-400/forced/hu) | [15.11](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v2/common_voice_17_0-argmax_subset-400/forced/ru) | [16.27](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v2/common_voice_17_0-argmax_subset-400/forced/gl) | [16.21](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v2/common_voice_17_0-argmax_subset-400/forced/fr) | [15.23](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v2/common_voice_17_0-argmax_subset-400/forced/pt) | [16.72](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v2/common_voice_17_0-argmax_subset-400/forced/da) | 3100 | [Link](https://github.com/argmaxinc/WhisperKit/commit/e3e21d4) |
61
- | [large-v3-v20240930](https://hf.co/argmaxinc/whisperkit-coreml/tree/main/openai_whisper-large-v3-v20240930) | [6.1](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930/common_voice_17_0-argmax_subset-400/forced/es) | [11.41](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930/common_voice_17_0-argmax_subset-400/forced/ro) | [23.3](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930/common_voice_17_0-argmax_subset-400/forced/th) | [8.91](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930/common_voice_17_0-argmax_subset-400/forced/nl) | [11.11](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930/common_voice_17_0-argmax_subset-400/forced/id) | [12.97](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930/common_voice_17_0-argmax_subset-400/forced/sv) | [12.26](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930/common_voice_17_0-argmax_subset-400/forced/de) | [12.12](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930/common_voice_17_0-argmax_subset-400/forced/pl) | [15.42](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930/common_voice_17_0-argmax_subset-400/forced/fi) | [12.83](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930/common_voice_17_0-argmax_subset-400/forced/it) | [12.85](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930/common_voice_17_0-argmax_subset-400/forced/cs) | [12.13](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930/common_voice_17_0-argmax_subset-400/forced/en) | [16.92](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930/common_voice_17_0-argmax_subset-400/forced/vi) | [17.73](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930/common_voice_17_0-argmax_subset-400/forced/el) | [15.3](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930/common_voice_17_0-argmax_subset-400/forced/hu) | [13.28](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930/common_voice_17_0-argmax_subset-400/forced/ru) | [15.0](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930/common_voice_17_0-argmax_subset-400/forced/gl) | [15.51](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930/common_voice_17_0-argmax_subset-400/forced/fr) | [14.93](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930/common_voice_17_0-argmax_subset-400/forced/pt) | [17.63](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930/common_voice_17_0-argmax_subset-400/forced/da) | 1640 | [Link](https://github.com/argmaxinc/WhisperKit/commit/e3e21d4) |
62
- | [large-v3-v20240930_626MB](https://hf.co/argmaxinc/whisperkit-coreml/tree/main/openai_whisper-large-v3-v20240930_626MB) | [5.97](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930_626MB/common_voice_17_0-argmax_subset-400/forced/es) | [12.24](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930_626MB/common_voice_17_0-argmax_subset-400/forced/ro) | [23.09](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930_626MB/common_voice_17_0-argmax_subset-400/forced/th) | [9.05](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930_626MB/common_voice_17_0-argmax_subset-400/forced/nl) | [12.66](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930_626MB/common_voice_17_0-argmax_subset-400/forced/id) | [12.72](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930_626MB/common_voice_17_0-argmax_subset-400/forced/sv) | [13.21](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930_626MB/common_voice_17_0-argmax_subset-400/forced/de) | [13.11](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930_626MB/common_voice_17_0-argmax_subset-400/forced/pl) | [15.17](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930_626MB/common_voice_17_0-argmax_subset-400/forced/fi) | [13.16](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930_626MB/common_voice_17_0-argmax_subset-400/forced/it) | [14.49](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930_626MB/common_voice_17_0-argmax_subset-400/forced/cs) | [13.03](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930_626MB/common_voice_17_0-argmax_subset-400/forced/en) | [17.36](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930_626MB/common_voice_17_0-argmax_subset-400/forced/vi) | [18.71](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930_626MB/common_voice_17_0-argmax_subset-400/forced/el) | [17.05](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930_626MB/common_voice_17_0-argmax_subset-400/forced/hu) | [14.37](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930_626MB/common_voice_17_0-argmax_subset-400/forced/ru) | [15.48](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930_626MB/common_voice_17_0-argmax_subset-400/forced/gl) | [15.68](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930_626MB/common_voice_17_0-argmax_subset-400/forced/fr) | [14.85](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930_626MB/common_voice_17_0-argmax_subset-400/forced/pt) | [18.94](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930_626MB/common_voice_17_0-argmax_subset-400/forced/da) | 626 | [Link](https://github.com/argmaxinc/WhisperKit/commit/e3e21d4) |
63
- | [large-v3-v20240930_547MB](https://hf.co/argmaxinc/whisperkit-coreml/tree/main/openai_whisper-large-v3-v20240930_547MB) | [7.84](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930_547MB/common_voice_17_0-argmax_subset-400/forced/es) | [18.26](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930_547MB/common_voice_17_0-argmax_subset-400/forced/ro) | [39.58](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930_547MB/common_voice_17_0-argmax_subset-400/forced/th) | [14.18](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930_547MB/common_voice_17_0-argmax_subset-400/forced/nl) | [17.25](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930_547MB/common_voice_17_0-argmax_subset-400/forced/id) | [19.25](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930_547MB/common_voice_17_0-argmax_subset-400/forced/sv) | [17.62](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930_547MB/common_voice_17_0-argmax_subset-400/forced/de) | [19.6](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930_547MB/common_voice_17_0-argmax_subset-400/forced/pl) | [20.31](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930_547MB/common_voice_17_0-argmax_subset-400/forced/fi) | [18.77](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930_547MB/common_voice_17_0-argmax_subset-400/forced/it) | [23.73](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930_547MB/common_voice_17_0-argmax_subset-400/forced/cs) | [16.12](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930_547MB/common_voice_17_0-argmax_subset-400/forced/en) | [25.97](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930_547MB/common_voice_17_0-argmax_subset-400/forced/vi) | [26.23](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930_547MB/common_voice_17_0-argmax_subset-400/forced/el) | [27.11](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930_547MB/common_voice_17_0-argmax_subset-400/forced/hu) | [18.63](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930_547MB/common_voice_17_0-argmax_subset-400/forced/ru) | [20.54](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930_547MB/common_voice_17_0-argmax_subset-400/forced/gl) | [22.0](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930_547MB/common_voice_17_0-argmax_subset-400/forced/fr) | [18.91](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930_547MB/common_voice_17_0-argmax_subset-400/forced/pt) | [25.3](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-large-v3-v20240930_547MB/common_voice_17_0-argmax_subset-400/forced/da) | 547 | [Link](https://github.com/argmaxinc/WhisperKit/commit/e3e21d4) |
64
- | [small](https://hf.co/argmaxinc/whisperkit-coreml/tree/main/openai_whisper-small) | [11.94](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-small/common_voice_17_0-argmax_subset-400/forced/es) | [26.99](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-small/common_voice_17_0-argmax_subset-400/forced/ro) | [21.52](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-small/common_voice_17_0-argmax_subset-400/forced/th) | [19.94](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-small/common_voice_17_0-argmax_subset-400/forced/nl) | [23.81](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-small/common_voice_17_0-argmax_subset-400/forced/id) | [23.97](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-small/common_voice_17_0-argmax_subset-400/forced/sv) | [23.87](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-small/common_voice_17_0-argmax_subset-400/forced/de) | [23.74](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-small/common_voice_17_0-argmax_subset-400/forced/pl) | [30.07](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-small/common_voice_17_0-argmax_subset-400/forced/fi) | [25.02](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-small/common_voice_17_0-argmax_subset-400/forced/it) | [37.7](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-small/common_voice_17_0-argmax_subset-400/forced/cs) | [17.35](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-small/common_voice_17_0-argmax_subset-400/forced/en) | [25.43](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-small/common_voice_17_0-argmax_subset-400/forced/vi) | [31.49](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-small/common_voice_17_0-argmax_subset-400/forced/el) | [44.66](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-small/common_voice_17_0-argmax_subset-400/forced/hu) | [26.09](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-small/common_voice_17_0-argmax_subset-400/forced/ru) | [30.45](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-small/common_voice_17_0-argmax_subset-400/forced/gl) | [27.11](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-small/common_voice_17_0-argmax_subset-400/forced/fr) | [35.7](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-small/common_voice_17_0-argmax_subset-400/forced/pt) | [37.18](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-small/common_voice_17_0-argmax_subset-400/forced/da) | 483 | [Link](https://github.com/argmaxinc/WhisperKit/commit/e3e21d4) |
65
- | [base](https://hf.co/argmaxinc/whisperkit-coreml/tree/main/openai_whisper-base) | [24.55](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-base/common_voice_17_0-argmax_subset-400/forced/es) | [54.19](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-base/common_voice_17_0-argmax_subset-400/forced/ro) | [32.91](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-base/common_voice_17_0-argmax_subset-400/forced/th) | [37.01](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-base/common_voice_17_0-argmax_subset-400/forced/nl) | [43.04](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-base/common_voice_17_0-argmax_subset-400/forced/id) | [45.53](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-base/common_voice_17_0-argmax_subset-400/forced/sv) | [38.09](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-base/common_voice_17_0-argmax_subset-400/forced/de) | [43.44](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-base/common_voice_17_0-argmax_subset-400/forced/pl) | [56.32](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-base/common_voice_17_0-argmax_subset-400/forced/fi) | [46.45](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-base/common_voice_17_0-argmax_subset-400/forced/it) | [67.24](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-base/common_voice_17_0-argmax_subset-400/forced/cs) | [25.11](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-base/common_voice_17_0-argmax_subset-400/forced/en) | [40.15](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-base/common_voice_17_0-argmax_subset-400/forced/vi) | [55.22](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-base/common_voice_17_0-argmax_subset-400/forced/el) | [71.07](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-base/common_voice_17_0-argmax_subset-400/forced/hu) | [44.21](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-base/common_voice_17_0-argmax_subset-400/forced/ru) | [47.63](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-base/common_voice_17_0-argmax_subset-400/forced/gl) | [45.09](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-base/common_voice_17_0-argmax_subset-400/forced/fr) | [48.98](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-base/common_voice_17_0-argmax_subset-400/forced/pt) | [61.96](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-base/common_voice_17_0-argmax_subset-400/forced/da) | 145 | [Link](https://github.com/argmaxinc/WhisperKit/commit/e3e21d4) |
66
- | [tiny](https://hf.co/argmaxinc/whisperkit-coreml/tree/main/openai_whisper-tiny) | [34.67](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-tiny/common_voice_17_0-argmax_subset-400/forced/es) | [66.78](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-tiny/common_voice_17_0-argmax_subset-400/forced/ro) | [41.88](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-tiny/common_voice_17_0-argmax_subset-400/forced/th) | [54.03](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-tiny/common_voice_17_0-argmax_subset-400/forced/nl) | [54.31](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-tiny/common_voice_17_0-argmax_subset-400/forced/id) | [64.66](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-tiny/common_voice_17_0-argmax_subset-400/forced/sv) | [49.11](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-tiny/common_voice_17_0-argmax_subset-400/forced/de) | [56.38](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-tiny/common_voice_17_0-argmax_subset-400/forced/pl) | [72.46](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-tiny/common_voice_17_0-argmax_subset-400/forced/fi) | [60.13](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-tiny/common_voice_17_0-argmax_subset-400/forced/it) | [81.53](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-tiny/common_voice_17_0-argmax_subset-400/forced/cs) | [33.47](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-tiny/common_voice_17_0-argmax_subset-400/forced/en) | [50.47](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-tiny/common_voice_17_0-argmax_subset-400/forced/vi) | [66.21](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-tiny/common_voice_17_0-argmax_subset-400/forced/el) | [85.67](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-tiny/common_voice_17_0-argmax_subset-400/forced/hu) | [59.73](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-tiny/common_voice_17_0-argmax_subset-400/forced/ru) | [54.05](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-tiny/common_voice_17_0-argmax_subset-400/forced/gl) | [59.49](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-tiny/common_voice_17_0-argmax_subset-400/forced/fr) | [65.65](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-tiny/common_voice_17_0-argmax_subset-400/forced/pt) | [79.84](https://hf.co/datasets/argmaxinc/whisperkit-evals-multilingual/tree/main/WhisperKit/openai_whisper-tiny/common_voice_17_0-argmax_subset-400/forced/da) | 66 | [Link](https://github.com/argmaxinc/WhisperKit/commit/e3e21d4) |
67
-
68
-
69
- ### Explanation
70
-
71
- We believe that rigorously measuring the quality of inference is necessary for developers and
72
- enterprises to make informed decisions when opting to use optimized or compressed variants of
73
- any machine learning model in production. To contextualize `WhisperKit`, we take the following Whisper
74
- implementations and benchmark them using a consistent evaluation harness:
75
-
76
- Server-side:
77
- - `WhisperOpenAIAPI`: [OpenAI's Whisper API](https://platform.openai.com/docs/guides/speech-to-text)
78
-
79
- ($0.36 per hour of audio as of 02/29/24, 25MB file size limit per request)
80
-
81
- On-device:
82
- - `WhisperKit`: Argmax's implementation [[Eval Harness]](https://github.com/argmaxinc/whisperkittools/blob/main/whisperkit/pipelines.py#L100) [[Repo]](https://github.com/argmaxinc/WhisperKit)
83
- - `whisper.cpp`: A C++ implementation form ggerganov [[Eval Harness]](https://github.com/argmaxinc/whisperkittools/blob/main/whisperkit/pipelines.py#L212) [[Repo]](https://github.com/ggerganov/whisper.cpp)
84
- - `WhisperMLX`: A Python implementation from Apple MLX [[Eval Harness]](https://github.com/argmaxinc/whisperkittools/blob/main/whisperkit/pipelines.py#L338) [[Repo]](https://github.com/ml-explore/mlx-examples/blob/main/whisper/whisper/transcribe.py)
85
-
86
- (All on-device implementations are available for free under MIT license as of 03/19/2024)
87
-
88
- `WhisperOpenAIAPI` sets the reference and we assume that it is using the equivalent of [openai/whisper-large-v2](https://huggingface.co/openai/whisper-large-v2)
89
- in float16 precision along with additional undisclosed optimizations from OpenAI. In all measurements, we care primarily about per-example no-regressions (quantified as `qoi` below)
90
- which is a stricter metric compared to dataset average [Word Error RATE (WER)](https://en.wikipedia.org/wiki/Word_error_rate). A 100% `qoi` preserves perfect backwards-compatibility on the test distribution and avoids "perceived regressions", the phenomenon
91
- where per-example known behavior changes after a code/model update and causes divergence in downstream code or breaks the user experience itself (even if dataset averages might stay flat
92
- across updates). Pseudocode for `qoi`:
93
-
94
- ```python
95
- qoi = []
96
- for example in dataset:
97
- no_regression = wer(optimized_model(example)) <= wer(reference_model(example))
98
- qoi.append(no_regression)
99
- qoi = (sum(qoi) / len(qoi)) * 100.
100
- ```
101
-
102
- Note that the ordering of models with respect to `WER` does not necessarily match the ordering with respect to `QoI`. This is because the reference model gets assigned
103
- a QoI of 100% by definition. Any per-example regression by other implementations get penalized while per-example improvements are not rewarded. `QoI` (higher is better) matters
104
- where the production behavior is established by the reference results and the goal is to not regress when switching to an optimized or compressed model. On the other hand,
105
- `WER` (lower is better) matters when there is no established production behavior and one is picking the best quality versus model size trade off point.
106
-
107
- We anticipate developers that use Whisper (or similar models) in production to have their own Quality Assurance test sets and [whisperkittools](https://github.com/argmaxinc/whisperkittools) offers
108
- the tooling necessary to run the same measurements on such custom test sets, please see the [Model Evaluation on Custom Dataset]((https://github.com/argmaxinc/whisperkittools)) for details.
109
-
110
- ### Why are there so many Whisper versions?
111
- WhisperKit is an SDK for building speech-to-text features in apps across a wide range of Apple devices. We are working towards abstracting away the model versioning from the developer so WhisperKit
112
- "just works" by deploying the highest-quality model version that a particular device can execute. In the interim, we leave the choice to the developer by providing quality and size trade-offs.
113
-
114
-
115
- ### Datasets
116
- - [librispeech](https://huggingface.co/datasets/argmaxinc/librispeech): ~5 hours of short English audio clips, tests short-form transcription quality
117
- - [earnings22](https://huggingface.co/datasets/argmaxinc/earnings22): ~120 hours of English audio clips from earnings calls with various accents, tests long-form transcription quality
118
- - [common_voice_17_0-argmax_subset-400](https://huggingface.co/datasets/argmaxinc/common_voice_17_0-argmax_subset-400): Up to 400 samples per language of multilingual audio clips with corresponding text, testing transcription quality across diverse languages and accents.
119
-
120
- ### Reproducing Results
121
- Benchmark results on this page were automatically generated by [whisperkittools](https://github.com/argmaxinc/whisperkittools) using our cluster of Apple Silicon Macs as self-hosted runners on
122
- Github Actions. We periodically recompute these benchmarks as part of our CI pipeline. Due to [security concerns](https://docs.github.com/en/actions/security-guides/security-hardening-for-github-actions#hardening-for-self-hosted-runners),
123
- we are unable to open up the cluster to the public. However, any Apple Silicon Mac (even with 8GB RAM) can be used to
124
- run identical [evaluation jobs](#evaluation) locally. For reference, our M2 Ultra devices complete a `librispeech` + `openai/whisper-large-v3`
125
- evaluation in under 1 hour regardless of the Whisper implementation. Oldest Apple Silicon Macs should take less than 1 day to complete the same evaluation.
126
-
127
-
128
-
129
- ### Glossary
130
-
131
- - `_turbo`: Indicates the presence of additional optimizations (not compression) to unlock streaming transcription
132
- as described in our [Blog Post](https://www.takeargmax.com/blog/whisperkit).
133
-
134
- - `_*MB`: Indicates the presence of model compression. Instead of cluttering the filename with details like
135
- `_AudioEncoder-5.8bits_TextDecoder-6.1bits_QLoRA-rank=16`, we choose to summarize the compression spec as the
136
- resulting total file size since this is what matters to developers in production.
137
 
 
1
 
2
  ---
3
+ pretty_name: "WhisperKit"
4
  viewer: false
5
  library_name: whisperkit
6
  tags:
 
10
  - asr
11
  - quantized
12
  ---
13
+ # WhisperKit
14
 
15
+ WhisperKit is an on-device speech recognition framework for Apple Silicon:
16
+ https://github.com/argmaxinc/WhisperKit
17
 
18
+ For performance and accuracy benchmarks on real devices, please see:
19
+ https://huggingface.co/spaces/argmaxinc/whisperkit-benchmarks
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
20