whisper.cpp
whisper.cpp copied to clipboard
Published
20 hours ago •
ggerganov
ggerganov
UTF8 issue with command line parameters in Windows version
If I pass the file "Chinese audio (中文).mp3" to the windows command line version, it exits with an errors:
rem Here main.exe has been renamed to whisper.exe
C:\...\whisp>whisper.exe --model models\ggml-tiny.bin --language chinese "Chinese file (中文).mp3"
whisper_init_from_file: loading model from 'models\ggml-tiny.bin'
whisper_model_load: loading model
whisper_model_load: n_vocab = 51865
whisper_model_load: n_audio_ctx = 1500
whisper_model_load: n_audio_state = 384
whisper_model_load: n_audio_head = 6
whisper_model_load: n_audio_layer = 4
whisper_model_load: n_text_ctx = 448
whisper_model_load: n_text_state = 384
whisper_model_load: n_text_head = 6
whisper_model_load: n_text_layer = 4
whisper_model_load: n_mels = 80
whisper_model_load: f16 = 1
whisper_model_load: type = 1
whisper_model_load: mem required = 127.00 MB (+ 3.00 MB per decoder)
whisper_model_load: kv self size = 2.62 MB
whisper_model_load: kv cross size = 8.79 MB
whisper_model_load: adding 1608 extra tokens
whisper_model_load: model ctx = 73.58 MB
whisper_model_load: model size = 73.54 MB
error: failed to open 'Chinese file (??).mp3' as WAV file
error: failed to read WAV file 'Chinese file (??).mp3'
whisper_print_timings: fallbacks = 0 p / 0 h
whisper_print_timings: load time = 398.52 ms
whisper_print_timings: mel time = 0.00 ms
whisper_print_timings: sample time = 0.00 ms / 1 runs ( 0.00 ms per run)
whisper_print_timings: encode time = 0.00 ms / 1 runs ( 0.00 ms per run)
whisper_print_timings: decode time = 0.00 ms / 1 runs ( 0.00 ms per run)
whisper_print_timings: total time = 399.54 ms
Runs fine when I rename the file omitting the chinese logograms.
I've also tried setting the codepage to UTF-8 with chcp 65001 with no luck.
(MacOS version works fine)
whisper.cppdoes not support.mp3files. The input has to be 16 kHz WAV
Yes, apologies, I copied the wrong output example.
This one is with a 16KHz wav file.
The problem are IMHO unicode characters in the file name:
D:\Tools\WhisperGUI>bin\whisper.exe --model models\ggml-tiny.en.bin --language en ..\file_中文.wav
whisper_init_from_file: loading model from 'models\ggml-tiny.en.bin'
whisper_model_load: loading model
whisper_model_load: n_vocab = 51864
whisper_model_load: n_audio_ctx = 1500
whisper_model_load: n_audio_state = 384
whisper_model_load: n_audio_head = 6
whisper_model_load: n_audio_layer = 4
whisper_model_load: n_text_ctx = 448
whisper_model_load: n_text_state = 384
whisper_model_load: n_text_head = 6
whisper_model_load: n_text_layer = 4
whisper_model_load: n_mels = 80
whisper_model_load: f16 = 1
whisper_model_load: type = 1
whisper_model_load: mem required = 127.00 MB (+ 3.00 MB per decoder)
whisper_model_load: kv self size = 2.62 MB
whisper_model_load: kv cross size = 8.79 MB
whisper_model_load: adding 1607 extra tokens
whisper_model_load: model ctx = 73.58 MB
whisper_model_load: model size = 73.54 MB
error: failed to open '..\file_??.wav' as WAV file
error: failed to read WAV file '..\file_??.wav'
whisper_print_timings: fallbacks = 0 p / 0 h
whisper_print_timings: load time = 494.26 ms
whisper_print_timings: mel time = 0.00 ms
whisper_print_timings: sample time = 0.00 ms / 1 runs ( 0.00 ms per run)
whisper_print_timings: encode time = 0.00 ms / 1 runs ( 0.00 ms per run)
whisper_print_timings: decode time = 0.00 ms / 1 runs ( 0.00 ms per run)
whisper_print_timings: total time = 495.03 ms
D:\Tools\WhisperGUI>copy ..\file_中文.wav ..\file.wav
1 file copiati.
D:\Tools\WhisperGUI>bin\whisper.exe --model models\ggml-tiny.en.bin --language en ..\file.wav
whisper_init_from_file: loading model from 'models\ggml-tiny.en.bin'
whisper_model_load: loading model
whisper_model_load: n_vocab = 51864
whisper_model_load: n_audio_ctx = 1500
whisper_model_load: n_audio_state = 384
whisper_model_load: n_audio_head = 6
whisper_model_load: n_audio_layer = 4
whisper_model_load: n_text_ctx = 448
whisper_model_load: n_text_state = 384
whisper_model_load: n_text_head = 6
whisper_model_load: n_text_layer = 4
whisper_model_load: n_mels = 80
whisper_model_load: f16 = 1
whisper_model_load: type = 1
whisper_model_load: mem required = 127.00 MB (+ 3.00 MB per decoder)
whisper_model_load: kv self size = 2.62 MB
whisper_model_load: kv cross size = 8.79 MB
whisper_model_load: adding 1607 extra tokens
whisper_model_load: model ctx = 73.58 MB
whisper_model_load: model size = 73.54 MB
system_info: n_threads = 4 / 8 | AVX = 1 | AVX2 = 1 | AVX512 = 0 | FMA = 0 | NEON = 0 | ARM_FMA = 0 | F16C = 0 | FP16_VA = 0 | WASM_SIMD = 0 | BLAS = 1 | SSE3 = 0 | VSX = 0 |
main: processing '..\file.wav' (63793110 samples, 3987.1 sec), 4 threads, 1 processors, lang = en, task = transcribe, timestamps = 1 ...
[00:00:00.000 --> 00:00:04.000] [speaking in foreign language]
[00:00:04.000 --> 00:00:08.000] [speaking in foreign language]
^C
This happens for the windows version only.
While the output on CMD terminal is correct, if you run chcp 65001, the bug is present (maybe just in parsing argv?) with --output-file, also:
D:\Tools\WhisperGUI>bin\main.exe --language chinese --model "models\ggml-tiny.bin" --output-file "D:\Lavori\result (中文)" -osrt -ovtt -otxt "D:\Lavori\5min.wav"
whisper_init_from_file: loading model from 'models\ggml-tiny.bin'
whisper_model_load: loading model
whisper_model_load: n_vocab = 51865
whisper_model_load: n_audio_ctx = 1500
whisper_model_load: n_audio_state = 384
whisper_model_load: n_audio_head = 6
whisper_model_load: n_audio_layer = 4
whisper_model_load: n_text_ctx = 448
whisper_model_load: n_text_state = 384
whisper_model_load: n_text_head = 6
whisper_model_load: n_text_layer = 4
whisper_model_load: n_mels = 80
whisper_model_load: f16 = 1
whisper_model_load: type = 1
whisper_model_load: mem required = 127.00 MB (+ 3.00 MB per decoder)
whisper_model_load: kv self size = 2.62 MB
whisper_model_load: kv cross size = 8.79 MB
whisper_model_load: adding 1608 extra tokens
whisper_model_load: model ctx = 73.58 MB
whisper_model_load: model size = 73.54 MB
system_info: n_threads = 4 / 8 | AVX = 1 | AVX2 = 1 | AVX512 = 0 | FMA = 0 | NEON = 0 | ARM_FMA = 0 | F16C = 0 | FP16_VA = 0 | WASM_SIMD = 0 | BLAS = 1 | SSE3 = 0 | VSX = 0 |
main: processing 'D:\Lavori\5min.wav' (4709599 samples, 294.3 sec), 4 threads, 1 processors, lang = chinese, task = transcribe, timestamps = 1 ...
[00:00:00.000 --> 00:00:02.000] 我現在回想
[00:00:02.000 --> 00:00:04.000] 我做信用電話
[00:00:04.000 --> 00:00:06.000] 事大美說是畢業以後
[00:00:06.000 --> 00:00:07.000] 我也錯過
[00:00:07.000 --> 00:00:10.000] 可是那個時候很自然人
- snip -
[00:04:44.080 --> 00:04:47.080] 这么多年来原门酒店天下
[00:04:47.080 --> 00:04:49.080] 加统夫罚但都问
[00:04:49.080 --> 00:04:51.080] 这么漂亮的字是谁写的
[00:04:51.080 --> 00:04:53.080] 是
[00:04:53.080 --> 00:04:54.280] 312歲的
output_txt: failed to open 'D:\Lavori\result (??).txt' for writing
output_vtt: failed to open 'D:\Lavori\result (??).vtt' for writing
output_srt: failed to open 'D:\Lavori\result (??).srt' for writing
whisper_print_timings: fallbacks = 17 p / 33 h
whisper_print_timings: load time = 486.94 ms
whisper_print_timings: mel time = 2188.15 ms
whisper_print_timings: sample time = 15614.09 ms / 7737 runs ( 2.02 ms per run)
whisper_print_timings: encode time = 14130.27 ms / 14 runs ( 1009.31 ms per run)
whisper_print_timings: decode time = 40749.71 ms / 7685 runs ( 5.30 ms per run)
whisper_print_timings: total time = 73315.52 ms
hmmmmm
even with chcp 65001
it still get bugged
shenjack whisper-bin-x64 ➜ ( master) ♥ 23:26 .\main.exe -m .\ggml-small.bin -f .\001.wav -t 12 -ocsv -of 001-small-cn -l auto -pp
whisper_init_from_file_no_state: loading model from '.\ggml-small.bin'
whisper_model_load: loading model
whisper_model_load: n_vocab = 51865
whisper_model_load: n_audio_ctx = 1500
whisper_model_load: n_audio_state = 768
whisper_model_load: n_audio_head = 12
whisper_model_load: n_audio_layer = 12
whisper_model_load: n_text_ctx = 448
whisper_model_load: n_text_state = 768
whisper_model_load: n_text_head = 12
whisper_model_load: n_text_layer = 12
whisper_model_load: n_mels = 80
whisper_model_load: f16 = 1
whisper_model_load: type = 3
whisper_model_load: mem required = 608.00 MB (+ 16.00 MB per decoder)
whisper_model_load: adding 1608 extra tokens
whisper_model_load: model ctx = 464.56 MB
whisper_model_load: model size = 464.44 MB
whisper_init_state: kv self size = 15.75 MB
whisper_init_state: kv cross size = 52.73 MB
system_info: n_threads = 12 / 12 | AVX = 1 | AVX2 = 1 | AVX512 = 0 | FMA = 1 | NEON = 0 | ARM_FMA = 0 | F16C = 1 | FP16_VA = 0 | WASM_SIMD = 0 | BLAS = 0 | SSE3 = 1 | VSX = 0 | COREML = 0 |
main: processing '.\001.wav' (870741 samples, 54.4 sec), 12 threads, 1 processors, lang = auto, task = transcribe, timestamps = 1 ...
whisper_full_with_state: auto-detected language: zh (p = 0.964870)
[00:00:00.000 --> 00:00:01.800] 涓夋閮芥槸鍦ㄥ鏍′腑涓
[00:00:01.800 --> 00:00:06.000] 鎵嶆渻鏈変簡灏嶆柤瀛哥敓楂瀷绠＄悊鐨勯€欏€嬫浠剁殑鏀
[00:00:06.000 --> 00:00:08.400] 灏辨槸鍦ㄤ笂涓婂眴鍓涚暍妤殑
[00:00:08.400 --> 00:00:10.600] 浠栧€戞妸鐢风敓鐨勯鍨嬫婧
[00:00:10.600 --> 00:00:13.000] 鐢变笁鍏垎瑾垮埌浜嗗叚鍏垎
[00:00:13.000 --> 00:00:16.000] 瑕佷笉鐒跺氨鎴戝啀娆℃斁涓€缍插ソ澶氫汉璁€鏂囬潻
[00:00:16.000 --> 00:00:17.600] 鍗充娇鐝惧湪鏄畝鏂囬潻涔熶笉鍚堟牸
[00:00:17.600 --> 00:00:19.000] 鐒℃剰寰屽洖浜嗙Ξ鎷
[00:00:19.000 --> 00:00:21.000] 閭ｆ垜瀹ｅ竷涓€涓嬮€欏牬姣旇辰绲愭灉
[00:00:21.000 --> 00:00:26.600] 鐛插緱鏈€浣宠畩鎴愮殑鏄鏂逛簩璁婂叏绁
whisper_full_with_state: progress = 5%
whisper_full_with_state: progress = 10%
whisper_full_with_state: progress = 15%
whisper_full_with_state: progress = 20%
whisper_full_with_state: progress = 25%
whisper_full_with_state: progress = 30%
whisper_full_with_state: progress = 35%
whisper_full_with_state: progress = 40%
whisper_full_with_state: progress = 45%
[00:00:26.600 --> 00:00:31.600] (鎺岃伈)
[00:00:31.600 --> 00:00:43.000] 鏈€寰岀嵅鍕濈殑鏄鏂
[00:00:43.000 --> 00:00:50.600] (鎺岃伈)
[00:00:50.600 --> 00:00:52.200] 閭ｇ従鍦ㄦ槸涓嶆槸鏂囬潻
[00:00:52.200 --> 00:00:53.400] 鏂囬潻
[00:00:53.400 --> 00:00:54.400] 閭ｅ挶鍊戞墦
whisper_full_with_state: progress = 50%
whisper_full_with_state: progress = 55%
whisper_full_with_state: progress = 60%
whisper_full_with_state: progress = 65%
whisper_full_with_state: progress = 70%
whisper_full_with_state: progress = 75%
whisper_full_with_state: progress = 80%
whisper_full_with_state: progress = 85%
whisper_full_with_state: progress = 90%
whisper_full_with_state: progress = 95%
output_csv: saving output to '001-small-cn.csv'
whisper_print_timings: load time = 360.42 ms
whisper_print_timings: fallbacks = 0 p / 0 h
whisper_print_timings: mel time = 132.76 ms
whisper_print_timings: sample time = 107.02 ms / 186 runs ( 0.58 ms per run)
whisper_print_timings: encode time = 8239.76 ms / 3 runs ( 2746.59 ms per run)
whisper_print_timings: decode time = 7407.09 ms / 187 runs ( 39.61 ms per run)
whisper_print_timings: total time = 16274.10 ms
shenjack whisper-bin-x64 ➜ ( master) ♥ 23:28 chcp
活动代码页: 65001
hmmmmm even with
chcp 65001it still get buggedshenjack whisper-bin-x64 ➜ ( master) ♥ 23:26 .\main.exe -m .\ggml-small.bin -f .\001.wav -t 12 -ocsv -of 001-small-cn -l auto -pp whisper_init_from_file_no_state: loading model from '.\ggml-small.bin' whisper_model_load: loading model whisper_model_load: n_vocab = 51865 whisper_model_load: n_audio_ctx = 1500 whisper_model_load: n_audio_state = 768 whisper_model_load: n_audio_head = 12 whisper_model_load: n_audio_layer = 12 whisper_model_load: n_text_ctx = 448 whisper_model_load: n_text_state = 768 whisper_model_load: n_text_head = 12 whisper_model_load: n_text_layer = 12 whisper_model_load: n_mels = 80 whisper_model_load: f16 = 1 whisper_model_load: type = 3 whisper_model_load: mem required = 608.00 MB (+ 16.00 MB per decoder) whisper_model_load: adding 1608 extra tokens whisper_model_load: model ctx = 464.56 MB whisper_model_load: model size = 464.44 MB whisper_init_state: kv self size = 15.75 MB whisper_init_state: kv cross size = 52.73 MB system_info: n_threads = 12 / 12 | AVX = 1 | AVX2 = 1 | AVX512 = 0 | FMA = 1 | NEON = 0 | ARM_FMA = 0 | F16C = 1 | FP16_VA = 0 | WASM_SIMD = 0 | BLAS = 0 | SSE3 = 1 | VSX = 0 | COREML = 0 | main: processing '.\001.wav' (870741 samples, 54.4 sec), 12 threads, 1 processors, lang = auto, task = transcribe, timestamps = 1 ... whisper_full_with_state: auto-detected language: zh (p = 0.964870) [00:00:00.000 --> 00:00:01.800] 涓夋閮芥槸鍦ㄥ鏍′腑涓 [00:00:01.800 --> 00:00:06.000] 鎵嶆渻鏈変簡灏嶆柤瀛哥敓楂瀷绠＄悊鐨勯€欏€嬫浠剁殑鏀 [00:00:06.000 --> 00:00:08.400] 灏辨槸鍦ㄤ笂涓婂眴鍓涚暍妤殑 [00:00:08.400 --> 00:00:10.600] 浠栧€戞妸鐢风敓鐨勯鍨嬫婧 [00:00:10.600 --> 00:00:13.000] 鐢变笁鍏垎瑾垮埌浜嗗叚鍏垎 [00:00:13.000 --> 00:00:16.000] 瑕佷笉鐒跺氨鎴戝啀娆℃斁涓€缍插ソ澶氫汉璁€鏂囬潻 [00:00:16.000 --> 00:00:17.600] 鍗充娇鐝惧湪鏄畝鏂囬潻涔熶笉鍚堟牸 [00:00:17.600 --> 00:00:19.000] 鐒℃剰寰屽洖浜嗙Ξ鎷 [00:00:19.000 --> 00:00:21.000] 閭ｆ垜瀹ｅ竷涓€涓嬮€欏牬姣旇辰绲愭灉 [00:00:21.000 --> 00:00:26.600] 鐛插緱鏈€浣宠畩鎴愮殑鏄鏂逛簩璁婂叏绁 whisper_full_with_state: progress = 5% whisper_full_with_state: progress = 10% whisper_full_with_state: progress = 15% whisper_full_with_state: progress = 20% whisper_full_with_state: progress = 25% whisper_full_with_state: progress = 30% whisper_full_with_state: progress = 35% whisper_full_with_state: progress = 40% whisper_full_with_state: progress = 45% [00:00:26.600 --> 00:00:31.600] (鎺岃伈) [00:00:31.600 --> 00:00:43.000] 鏈€寰岀嵅鍕濈殑鏄鏂 [00:00:43.000 --> 00:00:50.600] (鎺岃伈) [00:00:50.600 --> 00:00:52.200] 閭ｇ従鍦ㄦ槸涓嶆槸鏂囬潻 [00:00:52.200 --> 00:00:53.400] 鏂囬潻 [00:00:53.400 --> 00:00:54.400] 閭ｅ挶鍊戞墦 whisper_full_with_state: progress = 50% whisper_full_with_state: progress = 55% whisper_full_with_state: progress = 60% whisper_full_with_state: progress = 65% whisper_full_with_state: progress = 70% whisper_full_with_state: progress = 75% whisper_full_with_state: progress = 80% whisper_full_with_state: progress = 85% whisper_full_with_state: progress = 90% whisper_full_with_state: progress = 95% output_csv: saving output to '001-small-cn.csv' whisper_print_timings: load time = 360.42 ms whisper_print_timings: fallbacks = 0 p / 0 h whisper_print_timings: mel time = 132.76 ms whisper_print_timings: sample time = 107.02 ms / 186 runs ( 0.58 ms per run) whisper_print_timings: encode time = 8239.76 ms / 3 runs ( 2746.59 ms per run) whisper_print_timings: decode time = 7407.09 ms / 187 runs ( 39.61 ms per run) whisper_print_timings: total time = 16274.10 ms shenjack whisper-bin-x64 ➜ ( master) ♥ 23:28 chcp 活动代码页: 65001
use base model, same issue
