20240202在WIN10下使用whisper.cpp
2024/2/2 14:15
【结论：在Windows10下，确认large模式识别7分钟中文视频，需要83.7284 seconds，需要大概1.5分钟！效率太差！】
83.7284/420=0.19935333333333333333333333333333
前提条件，可以通过技术手段上外网！^_
首先你要有一张NVIDIA的显卡，比如我用的PDD拼多多的二手GTX1080显卡。【并且极其可能是矿卡！】800￥
2、请正确安装好NVIDIA最新的545版本的驱动程序和CUDA、cuDNN。
2、安装Torch
3、配置whisper
识别得到的字幕chs.srt是繁体中文的，将来要想办法更换为简体中文的！
1
00:00:00,000 --> 00:00:01,400
前段時間有個巨石恆虎
2
00:00:01,400 --> 00:00:03,000
某某是男人最好的醫妹
3
00:00:03,000 --> 00:00:04,800
這裡的某某可以替換為減肥
4
00:00:04,800 --> 00:00:07,800
長髮 西裝 考研 速唱 永潔無間等等等等
https://github.com/Const-me/Whisper/releases
https://www.cnblogs.com/jike9527/p/17545484.html?share_token=5af4092d-5b67-4e52-8231-0ae220fd2185
https://www.cnblogs.com/jike9527/p/17545484.html
使用whisper批量生成字幕(whisper.cpp)
c:\>
c:\>git clone https://github.com/ggerganov/whisper.cpp
Cloning into 'whisper.cpp'...
remote: Enumerating objects: 6773, done.
remote: Counting objects: 100% (1995/1995), done.
remote: Compressing objects: 100% (275/275), done.
remote: Total 6773 (delta 1826), reused 1810 (delta 1714), pack-reused 4778
Receiving objects: 100% (6773/6773), 10.18 MiB | 6.55 MiB/s, done.
Resolving deltas: 100% (4368/4368), done.
c:\>cd whisper.cpp
c:\whisper.cpp>dir
驱动器 C 中的卷是 WIN10
卷的序列号是 9273-D6A8
c:\whisper.cpp 的目录
2024/02/02 14:20 <DIR> .
2024/02/02 14:20 <DIR> ..
2024/02/02 14:20 <DIR> .devops
2024/02/02 14:20 <DIR> .github
2024/02/02 14:20 863 .gitignore
2024/02/02 14:20 99 .gitmodules
2024/02/02 14:20 <DIR> bindings
2024/02/02 14:20 <DIR> cmake
2024/02/02 14:20 19,729 CMakeLists.txt
2024/02/02 14:20 <DIR> coreml
2024/02/02 14:20 <DIR> examples
2024/02/02 14:20 <DIR> extra
2024/02/02 14:20 32,539 ggml-alloc.c
2024/02/02 14:20 4,149 ggml-alloc.h
2024/02/02 14:20 5,996 ggml-backend-impl.h
2024/02/02 14:20 69,048 ggml-backend.c
2024/02/02 14:20 11,932 ggml-backend.h
2024/02/02 14:20 451,408 ggml-cuda.cu
2024/02/02 14:20 2,156 ggml-cuda.h
2024/02/02 14:20 7,813 ggml-impl.h
2024/02/02 14:20 2,425 ggml-metal.h
2024/02/02 14:20 152,813 ggml-metal.m
2024/02/02 14:20 231,753 ggml-metal.metal
2024/02/02 14:20 87,989 ggml-opencl.cpp
2024/02/02 14:20 1,422 ggml-opencl.h
2024/02/02 14:20 411,673 ggml-quants.c
2024/02/02 14:20 13,983 ggml-quants.h
2024/02/02 14:20 696,627 ggml.c
2024/02/02 14:20 87,399 ggml.h
2024/02/02 14:20 <DIR> grammars
2024/02/02 14:20 1,093 LICENSE
2024/02/02 14:20 15,341 Makefile
2024/02/02 14:20 <DIR> models
2024/02/02 14:20 <DIR> openvino
2024/02/02 14:20 1,835 Package.swift
2024/02/02 14:20 39,942 README.md
2024/02/02 14:20 <DIR> samples
2024/02/02 14:20 <DIR> spm-headers
2024/02/02 14:20 <DIR> tests
2024/02/02 14:20 239,648 whisper.cpp
2024/02/02 14:20 30,873 whisper.h
26 个文件 2,620,548 字节
15 个目录 128,119,971,840 可用字节
c:\whisper.cpp>
c:\whisper.cpp>
c:\whisper.cpp>
c:\whisper.cpp>cd models
c:\whisper.cpp\models>dir
驱动器 C 中的卷是 WIN10
卷的序列号是 9273-D6A8
c:\whisper.cpp\models 的目录
2024/02/02 14:20 <DIR> .
2024/02/02 14:20 <DIR> ..
2024/02/02 14:20 7 .gitignore
2024/02/02 14:20 4,980 convert-h5-to-coreml.py
2024/02/02 14:20 7,584 convert-h5-to-ggml.py
2024/02/02 14:20 10,955 convert-pt-to-ggml.py
2024/02/02 14:20 12,761 convert-whisper-to-coreml.py
2024/02/02 14:20 1,799 convert-whisper-to-openvino.py
2024/02/02 14:20 2,272 download-coreml-model.sh
2024/02/02 14:20 1,440 download-ggml-model.cmd
2024/02/02 14:20 3,039 download-ggml-model.sh
2024/02/02 14:20 575,451 for-tests-ggml-base.bin
2024/02/02 14:20 586,836 for-tests-ggml-base.en.bin
2024/02/02 14:20 575,451 for-tests-ggml-large.bin
2024/02/02 14:20 575,451 for-tests-ggml-medium.bin
2024/02/02 14:20 586,836 for-tests-ggml-medium.en.bin
2024/02/02 14:20 575,451 for-tests-ggml-small.bin
2024/02/02 14:20 586,836 for-tests-ggml-small.en.bin
2024/02/02 14:20 575,451 for-tests-ggml-tiny.bin
2024/02/02 14:20 586,836 for-tests-ggml-tiny.en.bin
2024/02/02 14:20 1,506 generate-coreml-interface.sh
2024/02/02 14:20 1,355 generate-coreml-model.sh
2024/02/02 14:20 3,711 ggml_to_pt.py
2024/02/02 14:20 42 openvino-conversion-requirements.txt
2024/02/02 14:20 5,615 README.md
23 个文件 5,281,665 字节
2 个目录 105,396,047,872 可用字节
c:\whisper.cpp\models>main.exe -f samples\jfk.wav
'main.exe' 不是内部或外部命令，也不是可运行的程序
或批处理文件。
c:\whisper.cpp\models>dir
驱动器 C 中的卷是 WIN10
卷的序列号是 9273-D6A8
c:\whisper.cpp\models 的目录
2024/02/02 14:23 <DIR> .
2024/02/02 14:23 <DIR> ..
2024/02/02 14:20 7 .gitignore
2024/02/02 14:20 4,980 convert-h5-to-coreml.py
2024/02/02 14:20 7,584 convert-h5-to-ggml.py
2024/02/02 14:20 10,955 convert-pt-to-ggml.py
2024/02/02 14:20 12,761 convert-whisper-to-coreml.py
2024/02/02 14:20 1,799 convert-whisper-to-openvino.py
2024/02/02 14:20 2,272 download-coreml-model.sh
2024/02/02 14:20 1,440 download-ggml-model.cmd
2024/02/02 14:20 3,039 download-ggml-model.sh
2024/02/02 14:20 575,451 for-tests-ggml-base.bin
2024/02/02 14:20 586,836 for-tests-ggml-base.en.bin
2024/02/02 14:20 575,451 for-tests-ggml-large.bin
2024/02/02 14:20 575,451 for-tests-ggml-medium.bin
2024/02/02 14:20 586,836 for-tests-ggml-medium.en.bin
2024/02/02 14:20 575,451 for-tests-ggml-small.bin
2024/02/02 14:20 586,836 for-tests-ggml-small.en.bin
2024/02/02 14:20 575,451 for-tests-ggml-tiny.bin
2024/02/02 14:20 586,836 for-tests-ggml-tiny.en.bin
2024/02/02 14:20 1,506 generate-coreml-interface.sh
2024/02/02 14:20 1,355 generate-coreml-model.sh
2024/02/02 13:23 37,922,638 ggml-base-encoder.mlmodelc.zip
2024/02/02 13:23 59,707,625 ggml-base-q5_1.bin
2024/02/02 13:24 147,951,465 ggml-base.bin
2024/02/02 13:24 37,950,917 ggml-base.en-encoder.mlmodelc.zip
2024/02/02 13:24 59,721,011 ggml-base.en-q5_1.bin
2024/02/02 13:24 147,964,211 ggml-base.en.bin
2024/02/02 13:30 1,177,529,527 ggml-large-v1-encoder.mlmodelc.zip
2024/02/02 13:35 3,094,623,691 ggml-large-v1.bin
2024/02/02 13:31 1,174,643,458 ggml-large-v2-encoder.mlmodelc.zip
2024/02/02 13:30 1,080,732,091 ggml-large-v2-q5_0.bin
2024/02/02 13:35 3,094,623,691 ggml-large-v2.bin
2024/02/02 13:31 1,175,711,232 ggml-large-v3-encoder.mlmodelc.zip
2024/02/02 13:32 1,081,140,203 ggml-large-v3-q5_0.bin
2024/02/02 13:35 3,095,033,483 ggml-large-v3.bin
2024/02/02 13:57 567,829,413 ggml-medium-encoder.mlmodelc.zip
2024/02/02 13:57 539,212,467 ggml-medium-q5_0.bin
2024/02/02 14:03 1,533,763,059 ggml-medium.bin
2024/02/02 13:59 566,993,085 ggml-medium.en-encoder.mlmodelc.zip
2024/02/02 13:59 539,225,533 ggml-medium.en-q5_0.bin
2024/02/02 14:04 1,533,774,781 ggml-medium.en.bin
2024/02/02 14:08 163,083,239 ggml-small-encoder.mlmodelc.zip
2024/02/02 14:07 190,085,487 ggml-small-q5_1.bin
2024/02/02 14:09 487,601,967 ggml-small.bin
2024/02/02 14:09 162,952,446 ggml-small.en-encoder.mlmodelc.zip
2024/02/02 14:09 190,098,681 ggml-small.en-q5_1.bin
2024/02/02 14:11 487,614,201 ggml-small.en.bin
2024/02/02 14:10 15,037,446 ggml-tiny-encoder.mlmodelc.zip
2024/02/02 14:10 32,152,673 ggml-tiny-q5_1.bin
2024/02/02 14:11 77,691,713 ggml-tiny.bin
2024/02/02 14:11 15,034,655 ggml-tiny.en-encoder.mlmodelc.zip
2024/02/02 14:11 32,166,155 ggml-tiny.en-q5_1.bin
2024/02/02 14:12 43,550,795 ggml-tiny.en-q8_0.bin
2024/02/02 14:12 77,704,715 ggml-tiny.en.bin
2024/02/02 14:20 3,711 ggml_to_pt.py
2024/02/02 13:23 1,477 gitattributes
2024/02/02 14:20 42 openvino-conversion-requirements.txt
2024/02/02 13:23 1,311 README.md
57 个文件 22,726,106,592 字节
2 个目录 105,396,191,232 可用字节
c:\whisper.cpp\models>cd ..
c:\whisper.cpp>dir
c:\whisper.cpp>
c:\whisper.cpp>
c:\whisper.cpp>main.exe -f samples\jfk.wav
Using GPU "NVIDIA GeForce GTX 1080", feature level 12.1, effective flags Wave32 | NoReshapedMatMul
Loaded MEL filters, 62.8 kb RAM
Loaded vocabulary, 51864 strings, 3050.6 kb RAM
Loaded 245 GPU tensors, 140.539 MB VRAM
Computed CPU base frequency: 2.29469 GHz
Loaded model from "models/ggml-base.en.bin" to VRAM
Created source reader from the file "samples\jfk.wav"
00:00:00.000 --\> 00:00:11.000\] And so my fellow Americans, ask not what your country can do for you, ask what you can do for your country. CPU Tasks LoadModel 577.635 milliseconds RunComplete 422.9 milliseconds Run 319.505 milliseconds Callbacks 5.4751 milliseconds, 2 calls, 2.73755 milliseconds average Spectrogram 52.7935 milliseconds, 3 calls, 17.5978 milliseconds average Sample 7.6473 milliseconds, 27 calls, 283.233 microseconds average Encode 188.011 milliseconds Decode 125.975 milliseconds DecodeStep 118.306 milliseconds, 27 calls, 4.38169 milliseconds average GPU Tasks LoadModel 249.459 milliseconds Run 231.117 milliseconds Encode 99.0044 milliseconds EncodeLayer 77.7554 milliseconds, 6 calls, 12.9592 milliseconds average Decode 132.112 milliseconds DecodeStep 132.103 milliseconds, 27 calls, 4.89271 milliseconds average DecodeLayer 87.4824 milliseconds, 162 calls, 540.015 microseconds average Compute Shaders mulMatTiled 63.4898 milliseconds, 60 calls, 1.05816 milliseconds average mulMatByRowTiled 50.9198 milliseconds, 1959 calls, 25.9928 microseconds average softMaxLong 27.5314 milliseconds, 27 calls, 1.01968 milliseconds average norm 12.3785 milliseconds, 526 calls, 23.5333 microseconds average addRepeatGelu 11.9749 milliseconds, 170 calls, 70.4406 microseconds average fmaRepeat1 7.652 milliseconds, 526 calls, 14.5475 microseconds average addRepeatEx 7.4319 milliseconds, 498 calls, 14.9235 microseconds average softMaxFixed 6.913 milliseconds, 168 calls, 41.1488 microseconds average copyConvert 5.397 milliseconds, 348 calls, 15.5086 microseconds average convolutionMain 5.3903 milliseconds convolutionMain2Fixed 5.2572 milliseconds copyTranspose 4.6246 milliseconds, 336 calls, 13.7637 microseconds average scaleInPlace 4.5107 milliseconds, 168 calls, 26.8494 microseconds average addRepeatScale 3.7607 milliseconds, 324 calls, 11.6071 microseconds average softMax 2.9733 milliseconds, 162 calls, 18.3537 microseconds average addRepeat 1.8574 milliseconds, 180 calls, 10.3189 microseconds average diagMaskInf 1.3711 milliseconds, 162 calls, 8.46358 microseconds average convolutionPrep1 439.3 microseconds, 2 calls, 219.65 microseconds average convolutionPrep2 229.4 microseconds, 2 calls, 114.7 microseconds average addRows 191.5 microseconds, 27 calls, 7.09259 microseconds average add 60.4 microseconds mulMatByScalar 29.7 microseconds, 6 calls, 4.95 microseconds average mulMatByRow 27.6 microseconds, 6 calls, 4.6 microseconds average Memory Usage Model 858.5 KB RAM, 140.539 MB VRAM Context 1.19063 MB RAM, 186.732 MB VRAM Total 2.02901 MB RAM, 327.271 MB VRAM c:\\whisper.cpp\>main.exe -l zh -osrt -m models/ggml-medium.bin chs.wav Using GPU "NVIDIA GeForce GTX 1080", feature level 12.1, effective flags Wave32 \| NoReshapedMatMul Loaded MEL filters, 62.8 kb RAM Loaded vocabulary, 51865 strings, 3037.1 kb RAM Loaded 947 GPU tensors, 1462.12 MB VRAM Computed CPU base frequency: 2.29469 GHz Loaded model from "models/ggml-medium.bin" to VRAM Created source reader from the file "chs.wav" \[00:00:00.000 --\> 00:00:01.400\] ????????????? \[00:00:01.400 --\> 00:00:03.000\] ???????????? \[00:00:03.000 --\> 00:00:04.800\] ????????????????? \[00:00:04.800 --\> 00:00:07.800\] ??? ?? ??? ?? ????????? \[00:00:07.800 --\> 00:00:09.200\] ??????????? \[00:00:09.200 --\> 00:00:12.000\] ?????????????????????? \[00:00:12.000 --\> 00:00:13.400\] ????????? \[00:00:13.400 --\> 00:00:14.400\] ??????? \[00:00:14.400 --\> 00:00:17.400\] ????????????????????????? \[00:00:17.400 --\> 00:00:20.000\] ????????????????????? \[00:00:20.000 --\> 00:00:21.600\] ??????????????? \[00:00:21.600 --\> 00:00:22.800\] ????????? \[00:00:22.800 --\> 00:00:24.400\] ????????????? \[00:00:24.400 --\> 00:00:29.600\] ?????????????????? ????????????????????? \[00:00:29.600 --\> 00:00:32.400\] ??????? ???????? ??? \[00:00:32.400 --\> 00:00:34.600\] ?????????????????? \[00:00:34.600 --\> 00:00:36.200\] ??????????? \[00:00:36.200 --\> 00:00:37.000\] ??? \[00:00:37.000 --\> 00:00:38.000\] ????? \[00:00:38.000 --\> 00:00:39.400\] ??????????? \[00:00:39.400 --\> 00:00:40.600\] ???????? \[00:00:40.600 --\> 00:00:41.800\] ????? ????? \[00:00:41.800 --\> 00:00:44.000\] ??????????????????? \[00:00:44.000 --\> 00:00:46.600\] ????????????????????????? \[00:00:46.600 --\> 00:00:49.600\] ??????????????????????? \[00:00:49.600 --\> 00:00:52.000\] ??????????????????? \[00:00:52.000 --\> 00:00:54.200\] ??????????????????? \[00:00:54.200 --\> 00:00:56.000\] ??????? ?????? \[00:00:56.000 --\> 00:00:58.000\] ??????????????????? \[00:00:58.000 --\> 00:01:00.000\] ?????????????? \[00:01:00.000 --\> 00:01:01.000\] ???????? \[00:01:01.000 --\> 00:01:02.600\] ??????????? \[00:01:02.600 --\> 00:01:04.800\] ????????????? ???????? \[00:01:04.800 --\> 00:01:07.000\] ??11 ?????????????????? \[00:01:07.000 --\> 00:01:10.000\] ?????????????????? ???????? \[00:01:10.000 --\> 00:01:13.200\] ???? ??????????????????296% \[00:01:13.200 --\> 00:01:16.000\] ????????????????????? \[00:01:16.000 --\> 00:01:20.000\] ??????11 ?????? ????????????7???????? \[00:01:20.000 --\> 00:01:21.000\] ????????? \[00:01:21.000 --\> 00:01:22.400\] ??????????? \[00:01:22.400 --\> 00:01:24.200\] ???? ???????? \[00:01:24.200 --\> 00:01:26.800\] ??????????????????????? \[00:01:26.800 --\> 00:01:28.400\] ???? ????????? \[00:01:28.400 --\> 00:01:29.800\] ?????????? \[00:01:29.800 --\> 00:01:31.800\] ?????????????? ???? \[00:01:31.800 --\> 00:01:33.400\] ?????????????? \[00:01:33.400 --\> 00:01:35.400\] ??????????????? \[00:01:35.400 --\> 00:01:37.600\] ??? ?????2198 \[00:01:37.600 --\> 00:01:40.600\] ????????? ??????699 \[00:01:40.600 --\> 00:01:42.200\] ?????? ??????? \[00:01:42.200 --\> 00:01:45.000\] 400?????? ?????????300? \[00:01:45.000 --\> 00:01:48.200\] ??????? ????????200??????????? \[00:01:48.200 --\> 00:01:51.600\] ????? ????????????Citywalk???? \[00:01:51.600 --\> 00:01:54.600\] ?????? ???????1000???? \[00:01:54.600 --\> 00:01:58.200\] ???????????????????????????? \[00:01:58.200 --\> 00:02:00.400\] ????????????????? \[00:02:00.400 --\> 00:02:02.200\] ????????????? \[00:02:02.200 --\> 00:02:05.000\] ??????????????????????? \[00:02:05.000 --\> 00:02:07.400\] ????????? ??????????? \[00:02:07.400 --\> 00:02:08.600\] ???????? \[00:02:08.600 --\> 00:02:10.000\] ?????????? \[00:02:10.000 --\> 00:02:13.400\] ???????????????????????? ????1?1??? \[00:02:13.400 --\> 00:02:15.800\] ??????????????? ????? \[00:02:15.800 --\> 00:02:18.200\] ?????????? ????????? \[00:02:18.200 --\> 00:02:20.600\] ???????????? ??????? \[00:02:20.600 --\> 00:02:22.400\] ?????????? ??? \[00:02:22.400 --\> 00:02:26.400\] ????????? ????? ???? ?????????? \[00:02:26.400 --\> 00:02:29.200\] ???????? ??????????????????? \[00:02:29.200 --\> 00:02:30.800\] ???????????? \[00:02:30.800 --\> 00:02:32.600\] ???? ??????? \[00:02:32.600 --\> 00:02:35.400\] ????????? ???????? \[00:02:35.400 --\> 00:02:38.600\] ????????????? ??????????? \[00:02:38.600 --\> 00:02:41.000\] ?????? ??????????? \[00:02:41.000 --\> 00:02:43.600\] ?????????1000? ??????? \[00:02:43.600 --\> 00:02:46.400\] 500???????? 200??????? \[00:02:46.400 --\> 00:02:48.400\] ?99 ?????????? \[00:02:48.400 --\> 00:02:50.800\] ???????????? ????????? \[00:02:50.800 --\> 00:02:53.800\] ???????GORTEX??????? ??3000?? \[00:02:53.800 --\> 00:02:56.200\] ??????????????????????? \[00:02:56.200 --\> 00:03:00.000\] ???????????GORTEX???????????4500 \[00:03:00.000 --\> 00:03:03.000\] ?????GORTEX ????????????? \[00:03:03.000 --\> 00:03:05.800\] ????? ??????????????????? \[00:03:05.800 --\> 00:03:08.000\] ???????? ????? ???? \[00:03:08.000 --\> 00:03:09.800\] ????????????????? \[00:03:09.800 --\> 00:03:11.800\] ???????????????????? \[00:03:11.800 --\> 00:03:14.200\] ???????? ???????????? \[00:03:14.200 --\> 00:03:17.000\] ???????????? ???????? \[00:03:17.000 --\> 00:03:20.000\] ??????????? ?????????? \[00:03:20.000 --\> 00:03:21.600\] ???????????? \[00:03:21.600 --\> 00:03:23.200\] ????????????? \[00:03:23.200 --\> 00:03:26.000\] ????????????????? ????????????? \[00:03:26.000 --\> 00:03:29.000\] ??????????? ????????? ????????? \[00:03:29.000 --\> 00:03:31.800\] ?????????? ?????????????? \[00:03:31.800 --\> 00:03:35.000\] ??????? ???????????????????? \[00:03:35.000 --\> 00:03:36.800\] ???????????? \[00:03:36.800 --\> 00:03:40.000\] ???? ???????????? ??? \[00:03:40.000 --\> 00:03:42.600\] ?????????? ??????????? \[00:03:42.600 --\> 00:03:46.000\] ?????????? ???????????? \[00:03:46.000 --\> 00:03:49.200\] ??????????????? ????????????? \[00:03:49.200 --\> 00:03:52.200\] ?????????? ?????????? \[00:03:52.200 --\> 00:03:55.000\] ???????????????? ????? \[00:03:55.000 --\> 00:03:58.000\] ???????????? ????????????? \[00:03:58.000 --\> 00:04:01.000\] ?????????????????????? ????? \[00:04:01.000 --\> 00:04:04.000\] ??????????????? ?????? \[00:04:04.000 --\> 00:04:06.600\] ??????? ??????????????? \[00:04:06.600 --\> 00:04:08.800\] ??????????????? \[00:04:08.800 --\> 00:04:12.000\] ?????????????????? ????????? \[00:04:12.000 --\> 00:04:13.600\] ?????????????? \[00:04:13.600 --\> 00:04:16.200\] ??????????? ?????????? \[00:04:16.200 --\> 00:04:18.400\] ???????? ??????? \[00:04:18.400 --\> 00:04:21.800\] ?? ?????? ?????????????? \[00:04:21.800 --\> 00:04:25.800\] ??????????????? ?????????????????? \[00:04:25.800 --\> 00:04:29.200\] ???????? ???????????????????? \[00:04:29.200 --\> 00:04:30.800\] ????????????????? \[00:04:30.800 --\> 00:04:33.400\] ?????????? ????????? \[00:04:33.400 --\> 00:04:36.200\] ??????? ???????????????? \[00:04:36.200 --\> 00:04:39.400\] ???????? ??????????????? \[00:04:39.400 --\> 00:04:41.200\] ?????????????? \[00:04:41.200 --\> 00:04:43.600\] ?????????? ????????? \[00:04:43.600 --\> 00:04:45.000\] ?????????? \[00:04:45.000 --\> 00:04:47.600\] ???????????????????? \[00:04:47.600 --\> 00:04:51.600\] ????????????? ????????? ??????? \[00:04:51.600 --\> 00:04:53.200\] ??????????? \[00:04:53.200 --\> 00:04:55.800\] ??? ?????????????????????? \[00:04:55.800 --\> 00:04:57.400\] ???????????????? \[00:04:57.400 --\> 00:04:59.800\] ????????????????????? \[00:04:59.800 --\> 00:05:03.000\] ?????????????? ??????????? \[00:05:03.000 --\> 00:05:04.800\] ????????????????? \[00:05:04.800 --\> 00:05:07.200\] ???????????? ?????????? \[00:05:07.200 --\> 00:05:09.400\] ???? ?????????????? \[00:05:09.400 --\> 00:05:11.600\] ?????????????????? \[00:05:11.600 --\> 00:05:14.800\] ???????????????? ??????????? \[00:05:14.800 --\> 00:05:16.400\] ???? ?????? \[00:05:16.400 --\> 00:05:18.800\] ????? ?????????????? \[00:05:18.800 --\> 00:05:20.800\] ??????????????? \[00:05:20.800 --\> 00:05:23.200\] ????????? ???????????? \[00:05:23.200 --\> 00:05:25.600\] ????????? ?????????????? \[00:05:25.600 --\> 00:05:29.800\] ?????? ????????????????????881? \[00:05:29.800 --\> 00:05:31.800\] ??????? ??2000? \[00:05:31.800 --\> 00:05:34.600\] ?????? ?????????????????? \[00:05:34.600 --\> 00:05:38.400\] ?????????8000????????? 2000??????? \[00:05:38.600 --\> 00:05:41.200\] ????????? ???????????? \[00:05:41.200 --\> 00:05:43.600\] ?????? ??? ???????? \[00:05:43.600 --\> 00:05:46.600\] ??2000??8000???????????????? \[00:05:46.600 --\> 00:05:49.600\] ??????????? ?2018?2021? \[00:05:49.600 --\> 00:05:52.200\] ?????4???????60%?? \[00:05:52.200 --\> 00:05:56.000\] ??5??? ?????????????20??????60?? \[00:05:56.000 --\> 00:05:59.200\] ?????????? ????????????????? \[00:05:59.200 --\> 00:06:02.200\] ???????????? ????????????????? \[00:06:02.200 --\> 00:06:05.200\] ?????????? ??????????????? \[00:06:05.200 --\> 00:06:09.600\] ??? ????????? ???????????????????? \[00:06:09.600 --\> 00:06:11.400\] ???????????? \[00:06:11.400 --\> 00:06:15.200\] ???? ?????????? ???????????????? \[00:06:15.200 --\> 00:06:17.800\] ???? ???????????????? \[00:06:17.800 --\> 00:06:20.600\] ?350????????????????? \[00:06:20.600 --\> 00:06:23.000\] ??????? ?????????? \[00:06:23.000 --\> 00:06:25.000\] ????????????????? \[00:06:25.000 --\> 00:06:27.400\] ??? ???????????OK \[00:06:27.400 --\> 00:06:29.600\] ????????????????????? \[00:06:29.600 --\> 00:06:31.800\] ??????????????????? \[00:06:31.800 --\> 00:06:36.600\] ???????????????? ??????????????????????? \[00:06:36.600 --\> 00:06:38.800\] ????????????????? \[00:06:38.800 --\> 00:06:41.400\] ??????????????????? \[00:06:41.400 --\> 00:06:44.200\] ?????????????????????????? \[00:06:44.200 --\> 00:06:46.800\] ???????????????????? \[00:06:46.800 --\> 00:06:48.800\] ???????????????? \[00:06:48.800 --\> 00:06:51.200\] ??????????????????? \[00:06:51.200 --\> 00:06:53.000\] ???????????????? \[00:06:53.000 --\> 00:06:56.000\] ????????????????????????? \[00:06:56.000 --\> 00:07:01.600\] ????????????IC????? ????? ?????? CPU Tasks LoadModel 1.43866 seconds RunComplete 83.7284 seconds Run 83.6255 seconds Callbacks 457.784 milliseconds, 187 calls, 2.44804 milliseconds average Spectrogram 1.21106 seconds, 90 calls, 13.4562 milliseconds average Sample 1.01043 seconds, 3535 calls, 285.836 microseconds average Encode 15.2296 seconds, 17 calls, 895.858 milliseconds average Decode 67.9228 seconds, 17 calls, 3.99546 seconds average DecodeStep 66.9103 seconds, 3535 calls, 18.928 milliseconds average GPU Tasks LoadModel 1.03839 seconds Run 83.4773 seconds Encode 15.3219 seconds, 17 calls, 901.288 milliseconds average EncodeLayer 13.0778 seconds, 408 calls, 32.0533 milliseconds average Decode 68.1554 seconds, 17 calls, 4.00914 seconds average DecodeStep 68.1535 seconds, 3535 calls, 19.2796 milliseconds average DecodeLayer 61.7764 seconds, 84840 calls, 728.152 microseconds average Compute Shaders mulMatByRowTiled 38.8209 seconds, 1016702 calls, 38.1831 microseconds average mulMatTiled 15.8527 seconds, 8993 calls, 1.76278 milliseconds average fmaRepeat1 3.71454 seconds, 258888 calls, 14.348 microseconds average addRepeatEx 3.43395 seconds, 255336 calls, 13.4487 microseconds average normFixed 3.29705 seconds, 258888 calls, 12.7354 microseconds average softMaxLong 2.62421 seconds, 3535 calls, 742.351 microseconds average copyConvert 2.6175 seconds, 171312 calls, 15.2791 microseconds average addRepeatScale 2.43674 seconds, 169680 calls, 14.3608 microseconds average copyTranspose 2.43484 seconds, 170496 calls, 14.2809 microseconds average softMaxFixed 1.78188 seconds, 85248 calls, 20.9023 microseconds average addRepeatGelu 1.39165 seconds, 85282 calls, 16.3182 microseconds average softMax 1.27396 seconds, 84840 calls, 15.0161 microseconds average scaleInPlace 1.00817 seconds, 85248 calls, 11.8264 microseconds average addRepeat 954.089 milliseconds, 86064 calls, 11.0858 microseconds average diagMaskInf 652.093 milliseconds, 84840 calls, 7.68616 microseconds average convolutionMain2Fixed 388.382 milliseconds, 17 calls, 22.846 milliseconds average convolutionMain 163.663 milliseconds, 17 calls, 9.62722 milliseconds average convolutionPrep1 24.0373 milliseconds, 34 calls, 706.979 microseconds average addRows 21.3709 milliseconds, 3535 calls, 6.04552 microseconds average convolutionPrep2 7.0976 milliseconds, 34 calls, 208.753 microseconds average add 1.8821 milliseconds, 17 calls, 110.712 microseconds average Memory Usage Model 877.966 KB RAM, 1.42785 GB VRAM Context 109.465 MB RAM, 785.219 MB VRAM Total 110.322 MB RAM, 2.19467 GB VRAM c:\\whisper.cpp\>    https://github.com/ggerganov/whisper.cpp/tree/master/models https://github.com/ggerganov/whisper.cpp ggerganov/whisper.cpp https://blog.csdn.net/aiyolo/article/details/129674728?share_token=2c48b804-37f6-43a8-9159-08b28147ad67 Whisper.cpp 编译使用 whisper.cpp 是牛人 ggerganov 对 openai 的 whisper 语音识别模型用 C++ 重新实现的项目，开源在 github 上，具有轻量、性能高，实用性强等特点。这篇文章主要记录在 windows 平台，如何使用该模型在本地端进行语音识别。 whisper.cpp 的开源地址在 ggerganov/whisper.cpp: Port of OpenAI's Whisper model in C/C++ (github.com)，首先将项目下载在本地。 git clone https://github.com/ggerganov/whisper.cpp whisper.cpp 项目里提供了几个现成的模型。建议下载 small 以上的模型，不然识别效果完全无法使用。 https://huggingface.co/ggerganov/whisper.cpp ggerganov/whisper.cpp OpenAI's Whisper models converted to ggml format Available models Model Disk Mem SHA tiny 75 MB \~390 MB bd577a113a864445d4c299885e0cb97d4ba92b5f tiny.en 75 MB \~390 MB c78c86eb1a8faa21b369bcd33207cc90d64ae9df base 142 MB \~500 MB 465707469ff3a37a2b9b8d8f89f2f99de7299dac base.en 142 MB \~500 MB 137c40403d78fd54d454da0f9bd998f78703390c small 466 MB \~1.0 GB 55356645c2b361a969dfd0ef2c5a50d530afd8d5 small.en 466 MB \~1.0 GB db8a495a91d927739e50b3fc1cc4c6b8f6c2d022 medium 1.5 GB \~2.6 GB fd9727b6e1217c2f614f9b698455c4ffd82463b4 medium.en 1.5 GB \~2.6 GB 8c30f0e44ce9560643ebd10bbe50cd20eafd3723 large-v1 2.9 GB \~4.7 GB b1caaf735c4cc1429223d5a74f0f4d0b9b59a299 large-v2 2.9 GB \~4.7 GB 0f4c8e34f21cf1a914c59d8b3ce882345ad349d6 large 2.9 GB \~4.7 GB ad82bf6a9043ceed055076d0fd39f5f186ff8062 note: large corresponds to the latest Large v3 model For more information, visit: https://github.com/ggerganov/whisper.cpp/tree/master/models https://huggingface.co/ggerganov/whisper.cpp/tree/main 参考资料： https://www.toutiao.com/article/7225218604160418338/?app=news_article\×tamp=1706803458\&use_new_style=1\&req_id=2024020200041726E9258609E554857D25\&group_id=7225218604160418338\&tt_from=mobile_qq\&utm_source=mobile_qq\&utm_medium=toutiao_android\&utm_campaign=client_share\&share_token=37e094d5-29b8-4d14-87bb-241cdc28b0ea\&source=m_redirect AI浪潮下的12大开源神器介绍 原创2023-04-23 20:33·IT小熊实验室丶 https://blog.csdn.net/sinat_18131557/article/details/130950719?share_token=25ca6bb5-8450-472c-9228-abc8c6ce74d8 whisper.cpp在Windows VS的编译 sinat_18131557 于 2023-05-30 16:03:53 发布 https://www.toutiao.com/article/7283079784329052726/?app=news_article\×tamp=1706803297\&use_new_style=1\&req_id=20240202000137411974769524167990E0\&group_id=7283079784329052726\&tt_from=mobile_qq\&utm_source=mobile_qq\&utm_medium=toutiao_android\&utm_campaign=client_share\&share_token=b7961b29-d87a-4b6c-bb8e-c7c213388390\&source=m_redirect 【往期回顾】Github开源项目月刊精选-2023年8月 原创2023-09-27 08:30·Github推荐官 https://blog.csdn.net/weixin_45533131/article/details/132817683?share_token=72d8a161-4d49-4795-ad21-2ce5e2e4b197 在Linux(Centos7)上编译whisper.cpp的详细教程 https://blog.csdn.net/u012234115/article/details/134668510?share_token=e3835a0d-ac3b-4c86-9e32-e79ec85cddbe 开源C++智能语音识别库whisper.cpp开发使用入门 https://www.toutiao.com/article/7276732434920653312/?app=news_article\×tamp=1706802934\&use_new_style=1\&req_id=2024020123553463D3509B1706BC79D479\&group_id=7276732434920653312\&tt_from=mobile_qq\&utm_source=mobile_qq\&utm_medium=toutiao_android\&utm_campaign=client_share\&share_token=7bcb7488-a03d-4291-96fb-d0835ac76cca\&source=m_redirect OpenAI的whisper的c/c++ 版本体验 首先下载代码,注：我的OS环境是ubuntu 18.04。 https://post.smzdm.com/p/a3052kz7/?share_token=d4057cba-adb0-4c91-8a8b-d8a7adcf4087 显卡怎么玩 篇三：音频转字幕神器whisper升级版，whisper-webui使用教程 https://www.toutiao.com/article/7311876528407921162/?app=news_article\×tamp=1706801102\&use_new_style=1\&req_id=20240201232501647517150775FC7AD89A\&group_id=7311876528407921162\&tt_from=mobile_qq\&utm_source=mobile_qq\&utm_medium=toutiao_android\&utm_campaign=client_share\&share_token=dfa1976e-9422-49d2-a73b-6453becea90c\&source=m_redirect 2023 AI 界7个最火的 Text-to-Video 模型 动画 https://www.toutiao.com/article/7312473532829745700/?app=news_article\×tamp=1706801052\&use_new_style=1\&req_id=2024020123241265D9BE3F954EB979A010\&group_id=7312473532829745700\&tt_from=mobile_qq\&utm_source=mobile_qq\&utm_medium=toutiao_android\&utm_campaign=client_share\&share_token=ca5d0d2a-2d9b-4959-b5c0-3dd869555240\&source=m_redirect 推荐5款本周 超火 的开源AI项目 原创2023-12-15 07:32·程序员梓羽同学 https://blog.csdn.net/chenlu5201314/article/details/131156770?share_token=b8796ff0-44f8-471a-af6d-c1bc7ca57002 【开源工具】使用Whisper提取视频、语音的字幕 1、下载安装包Assets\\WhisperDesktop.zip https://www.toutiao.com/article/7222852915286016544/?app=news_article\×tamp=1706460752\&use_new_style=1\&req_id=2024012900523164164830D4E1ECF3CCE2\&group_id=7222852915286016544\&tt_from=mobile_qq\&utm_source=mobile_qq\&utm_medium=toutiao_android\&utm_campaign=client_share\&share_token=9bc8621f-b3b1-4f49-ae20-5214c1254515\&source=m_redirect 从零开始，手把手教本地部署Stable Diffusion AI绘画 V3版 (Win最新) 原创2023-04-17 11:23·觉悟之坡 https://blog.csdn.net/S_eashell/article/details/135258411?share_token=f998e896-6dff-4fd4-8df2-c6aae132e95c 98秒转录2.5小时音频，最强音频转文字软件insanely-fast-whisper下载部署 老艾的AI世界 已于 2024-01-05 20:20:51 修改