20240131在ubuntu20.04.6下配置whisper

2024/1/31 15:48

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最新的驱动程序和CUDA。可选安装！

3、配置whisper

rootroot@rootroot-X99-Turbo:~$
rootroot@rootroot-X99-Turbo:~$ python -m pip install --upgrade pip
【可以不安装conda】
rootroot@rootroot-X99-Turbo:~$ wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh
rootroot@rootroot-X99-Turbo:~$ ffmpeg
rootroot@rootroot-X99-Turbo:~$ pip install -U openai-whisper
rootroot@rootroot-X99-Turbo:~$ pip install tiktoken
rootroot@rootroot-X99-Turbo:~$ pip install setuptools-rust
rootroot@rootroot-X99-Turbo:~$ whisper audio.mp3 --model medium --language Chinese
rootroot@rootroot-X99-Turbo:~$ whisper chi.mp4 --model medium --language Chinese
rootroot@rootroot-X99-Turbo:~$ sudo apt-get install ffmpeg
rootroot@rootroot-X99-Turbo:~$ time(whisper chs.mp4 --model medium --language Chinese)

rootroot@rootroot-X99-Turbo:~$

rootroot@rootroot-X99-Turbo:~$ python -m pip install --upgrade pip

Collecting pip

Downloading pip-23.3.2-py3-none-any.whl (2.1 MB)

|████████████████████████████████| 2.1 MB 690 kB/s

Installing collected packages: pip

Successfully installed pip-23.3.2

rootroot@rootroot-X99-Turbo:~$

rootroot@rootroot-X99-Turbo:~$ sudo mkdir /opt/tools

rootroot@rootroot-X99-Turbo:~$ cd /opt/tools/

rootroot@rootroot-X99-Turbo:/opt/tools$

rootroot@rootroot-X99-Turbo:/opt/tools$ ll

total 8

drwxr-xr-x 2 root root 4096 1月 26 12:21 ./

drwxr-xr-x 4 root root 4096 1月 26 12:21 ../

rootroot@rootroot-X99-Turbo:/opt/tools$

rootroot@rootroot-X99-Turbo:/opt/tools$ cd ~

rootroot@rootroot-X99-Turbo:~$

rootroot@rootroot-X99-Turbo:~$ wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh

--2024-01-26 12:22:28-- https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh

Resolving repo.anaconda.com (repo.anaconda.com)... 104.16.130.3, 104.16.131.3, 2606:4700::6810:8203, ...

Connecting to repo.anaconda.com (repo.anaconda.com)|104.16.130.3|:443... connected.

HTTP request sent, awaiting response... 200 OK

Length: 141613749 (135M) [application/octet-stream]

Saving to: 'Miniconda3-latest-Linux-x86_64.sh'

Miniconda3-latest-Linux-x86_64.sh 100%[=============================================================================================>] 135.05M 2.82MB/s in 51s

2024-01-26 12:23:20 (2.65 MB/s) - 'Miniconda3-latest-Linux-x86_64.sh' saved [141613749/141613749]

rootroot@rootroot-X99-Turbo:~$ ffmpeg

built with gcc 9 (Ubuntu 9.4.0-1ubuntu1~20.04.1)

configuration: --prefix=/usr --extra-version=0ubuntu0.1 --toolchain=hardened --libdir=/usr/lib/x86_64-linux-gnu --incdir=/usr/include/x86_64-linux-gnu --arch=amd64 --enable-gpl --disable-stripping --enable-avresample --disable-filter=resample --enable-avisynth --enable-gnutls --enable-ladspa --enable-libaom --enable-libass --enable-libbluray --enable-libbs2b --enable-libcaca --enable-libcdio --enable-libcodec2 --enable-libflite --enable-libfontconfig --enable-libfreetype --enable-libfribidi --enable-libgme --enable-libgsm --enable-libjack --enable-libmp3lame --enable-libmysofa --enable-libopenjpeg --enable-libopenmpt --enable-libopus --enable-libpulse --enable-librsvg --enable-librubberband --enable-libshine --enable-libsnappy --enable-libsoxr --enable-libspeex --enable-libssh --enable-libtheora --enable-libtwolame --enable-libvidstab --enable-libvorbis --enable-libvpx --enable-libwavpack --enable-libwebp --enable-libx265 --enable-libxml2 --enable-libxvid --enable-libzmq --enable-libzvbi --enable-lv2 --enable-omx --enable-openal --enable-opencl --enable-opengl --enable-sdl2 --enable-libdc1394 --enable-libdrm --enable-libiec61883 --enable-nvenc --enable-chromaprint --enable-frei0r --enable-libx264 --enable-shared

libavutil 56. 31.100 / 56. 31.100

libavcodec 58. 54.100 / 58. 54.100

libavformat 58. 29.100 / 58. 29.100

libavdevice 58. 8.100 / 58. 8.100

libavfilter 7. 57.100 / 7. 57.100

libavresample 4. 0. 0 / 4. 0. 0

libswscale 5. 5.100 / 5. 5.100

libswresample 3. 5.100 / 3. 5.100

libpostproc 55. 5.100 / 55. 5.100

Hyper fast Audio and Video encoder

usage: ffmpeg [options] [[infile options] -i infile]... {[outfile options] outfile}...

Use -h to get full help or, even better, run 'man ffmpeg'

rootroot@rootroot-X99-Turbo:~$

rootroot@rootroot-X99-Turbo:~$ pip install -U openai-whisper

Defaulting to user installation because normal site-packages is not writeable

Requirement already satisfied: openai-whisper in ./.local/lib/python3.8/site-packages (20231117)

Requirement already satisfied: triton<3,>=2.0.0 in ./.local/lib/python3.8/site-packages (from openai-whisper) (2.2.0)

Requirement already satisfied: numba in ./.local/lib/python3.8/site-packages (from openai-whisper) (0.58.1)

Requirement already satisfied: numpy in ./.local/lib/python3.8/site-packages (from openai-whisper) (1.24.4)

Requirement already satisfied: torch in ./.local/lib/python3.8/site-packages (from openai-whisper) (2.1.2)

Requirement already satisfied: tqdm in ./.local/lib/python3.8/site-packages (from openai-whisper) (4.66.1)

Requirement already satisfied: more-itertools in ./.local/lib/python3.8/site-packages (from openai-whisper) (10.2.0)

Requirement already satisfied: tiktoken in ./.local/lib/python3.8/site-packages (from openai-whisper) (0.5.2)

Requirement already satisfied: filelock in ./.local/lib/python3.8/site-packages (from triton<3,>=2.0.0->openai-whisper) (3.13.1)

Requirement already satisfied: llvmlite<0.42,>=0.41.0dev0 in ./.local/lib/python3.8/site-packages (from numba->openai-whisper) (0.41.1)

Requirement already satisfied: importlib-metadata in ./.local/lib/python3.8/site-packages (from numba->openai-whisper) (7.0.1)

Requirement already satisfied: regex>=2022.1.18 in ./.local/lib/python3.8/site-packages (from tiktoken->openai-whisper) (2023.12.25)

Requirement already satisfied: requests>=2.26.0 in ./.local/lib/python3.8/site-packages (from tiktoken->openai-whisper) (2.31.0)

Requirement already satisfied: typing-extensions in ./.local/lib/python3.8/site-packages (from torch->openai-whisper) (4.9.0)

Requirement already satisfied: sympy in ./.local/lib/python3.8/site-packages (from torch->openai-whisper) (1.12)

Requirement already satisfied: networkx in ./.local/lib/python3.8/site-packages (from torch->openai-whisper) (3.1)

Requirement already satisfied: jinja2 in ./.local/lib/python3.8/site-packages (from torch->openai-whisper) (3.1.3)

Requirement already satisfied: fsspec in ./.local/lib/python3.8/site-packages (from torch->openai-whisper) (2023.12.2)

Requirement already satisfied: nvidia-cuda-nvrtc-cu12==12.1.105 in ./.local/lib/python3.8/site-packages (from torch->openai-whisper) (12.1.105)

Requirement already satisfied: nvidia-cuda-runtime-cu12==12.1.105 in ./.local/lib/python3.8/site-packages (from torch->openai-whisper) (12.1.105)

Requirement already satisfied: nvidia-cuda-cupti-cu12==12.1.105 in ./.local/lib/python3.8/site-packages (from torch->openai-whisper) (12.1.105)

Requirement already satisfied: nvidia-cudnn-cu12==8.9.2.26 in ./.local/lib/python3.8/site-packages (from torch->openai-whisper) (8.9.2.26)

Requirement already satisfied: nvidia-cublas-cu12==12.1.3.1 in ./.local/lib/python3.8/site-packages (from torch->openai-whisper) (12.1.3.1)

Requirement already satisfied: nvidia-cufft-cu12==11.0.2.54 in ./.local/lib/python3.8/site-packages (from torch->openai-whisper) (11.0.2.54)

Requirement already satisfied: nvidia-curand-cu12==10.3.2.106 in ./.local/lib/python3.8/site-packages (from torch->openai-whisper) (10.3.2.106)

Requirement already satisfied: nvidia-cusolver-cu12==11.4.5.107 in ./.local/lib/python3.8/site-packages (from torch->openai-whisper) (11.4.5.107)

Requirement already satisfied: nvidia-cusparse-cu12==12.1.0.106 in ./.local/lib/python3.8/site-packages (from torch->openai-whisper) (12.1.0.106)

Requirement already satisfied: nvidia-nccl-cu12==2.18.1 in ./.local/lib/python3.8/site-packages (from torch->openai-whisper) (2.18.1)

Requirement already satisfied: nvidia-nvtx-cu12==12.1.105 in ./.local/lib/python3.8/site-packages (from torch->openai-whisper) (12.1.105)

Collecting triton<3,>=2.0.0 (from openai-whisper)

Downloading triton-2.1.0-0-cp38-cp38-manylinux2014_x86_64.manylinux_2_17_x86_64.whl.metadata (1.3 kB)

Requirement already satisfied: nvidia-nvjitlink-cu12 in ./.local/lib/python3.8/site-packages (from nvidia-cusolver-cu12==11.4.5.107->torch->openai-whisper) (12.3.101)

Requirement already satisfied: charset-normalizer<4,>=2 in ./.local/lib/python3.8/site-packages (from requests>=2.26.0->tiktoken->openai-whisper) (3.3.2)

Requirement already satisfied: idna<4,>=2.5 in /usr/lib/python3/dist-packages (from requests>=2.26.0->tiktoken->openai-whisper) (2.8)

Requirement already satisfied: urllib3<3,>=1.21.1 in /usr/lib/python3/dist-packages (from requests>=2.26.0->tiktoken->openai-whisper) (1.25.8)

Requirement already satisfied: certifi>=2017.4.17 in /usr/lib/python3/dist-packages (from requests>=2.26.0->tiktoken->openai-whisper) (2019.11.28)

Requirement already satisfied: zipp>=0.5 in ./.local/lib/python3.8/site-packages (from importlib-metadata->numba->openai-whisper) (3.17.0)

Requirement already satisfied: MarkupSafe>=2.0 in ./.local/lib/python3.8/site-packages (from jinja2->torch->openai-whisper) (2.1.3)

Requirement already satisfied: mpmath>=0.19 in ./.local/lib/python3.8/site-packages (from sympy->torch->openai-whisper) (1.3.0)

Downloading triton-2.1.0-0-cp38-cp38-manylinux2014_x86_64.manylinux_2_17_x86_64.whl (89.2 MB)

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 89.2/89.2 MB 25.9 MB/s eta 0:00:00

Installing collected packages: triton

Attempting uninstall: triton

Found existing installation: triton 2.2.0

Uninstalling triton-2.2.0:

Successfully uninstalled triton-2.2.0

Successfully installed triton-2.1.0

rootroot@rootroot-X99-Turbo:~$

rootroot@rootroot-X99-Turbo:~$ pip install tiktoken

Defaulting to user installation because normal site-packages is not writeable

Requirement already satisfied: tiktoken in ./.local/lib/python3.8/site-packages (0.5.2)

Requirement already satisfied: regex>=2022.1.18 in ./.local/lib/python3.8/site-packages (from tiktoken) (2023.12.25)

Requirement already satisfied: requests>=2.26.0 in ./.local/lib/python3.8/site-packages (from tiktoken) (2.31.0)

Requirement already satisfied: charset-normalizer<4,>=2 in ./.local/lib/python3.8/site-packages (from requests>=2.26.0->tiktoken) (3.3.2)

Requirement already satisfied: idna<4,>=2.5 in /usr/lib/python3/dist-packages (from requests>=2.26.0->tiktoken) (2.8)

Requirement already satisfied: urllib3<3,>=1.21.1 in /usr/lib/python3/dist-packages (from requests>=2.26.0->tiktoken) (1.25.8)

Requirement already satisfied: certifi>=2017.4.17 in /usr/lib/python3/dist-packages (from requests>=2.26.0->tiktoken) (2019.11.28)

rootroot@rootroot-X99-Turbo:~$

rootroot@rootroot-X99-Turbo:~$ pip install setuptools-rust

Defaulting to user installation because normal site-packages is not writeable

Requirement already satisfied: setuptools-rust in ./.local/lib/python3.8/site-packages (1.8.1)

Requirement already satisfied: setuptools>=62.4 in ./.local/lib/python3.8/site-packages (from setuptools-rust) (69.0.3)

Requirement already satisfied: semantic-version<3,>=2.8.2 in ./.local/lib/python3.8/site-packages (from setuptools-rust) (2.10.0)

Requirement already satisfied: tomli>=1.2.1 in ./.local/lib/python3.8/site-packages (from setuptools-rust) (2.0.1)

rootroot@rootroot-X99-Turbo:~$ sudo apt update && sudo apt install ffmpeg

Get:1 file:/var/cuda-repo-ubuntu2004-12-0-local InRelease [1,575 B]

Get:2 file:/var/cuda-repo-ubuntu2004-12-3-local InRelease [1,572 B]

Get:1 file:/var/cuda-repo-ubuntu2004-12-0-local InRelease [1,575 B]

Get:2 file:/var/cuda-repo-ubuntu2004-12-3-local InRelease [1,572 B]

Hit:3 http://mirrors.tuna.tsinghua.edu.cn/ubuntu focal InRelease

Hit:4 http://mirrors.tuna.tsinghua.edu.cn/ubuntu focal-updates InRelease

Hit:5 http://mirrors.tuna.tsinghua.edu.cn/ubuntu focal-backports InRelease

Hit:6 http://security.ubuntu.com/ubuntu focal-security InRelease

Hit:7 http://ppa.launchpad.net/graphics-drivers/ppa/ubuntu focal InRelease

Reading package lists... Done

Building dependency tree

Reading state information... Done

30 packages can be upgraded. Run 'apt list --upgradable' to see them.

Reading package lists... Done

Building dependency tree

Reading state information... Done

ffmpeg is already the newest version (7:4.2.7-0ubuntu0.1).

0 upgraded, 0 newly installed, 0 to remove and 30 not upgraded.

rootroot@rootroot-X99-Turbo:~$

rootroot@rootroot-X99-Turbo:~$ whisper audio.mp3 --model medium --language Chinese

100%|█████████████████████████████████████| 1.42G/1.42G [03:24<00:00, 7.48MiB/s]

Traceback (most recent call last):

File "/home/rootroot/.local/lib/python3.8/site-packages/whisper/audio.py", line 58, in load_audio

out = run(cmd, capture_output=True, check=True).stdout

File "/usr/lib/python3.8/subprocess.py", line 516, in run

raise CalledProcessError(retcode, process.args,

subprocess.CalledProcessError: Command '['ffmpeg', '-nostdin', '-threads', '0', '-i', 'audio.mp3', '-f', 's16le', '-ac', '1', '-acodec', 'pcm_s16le', '-ar', '16000', '-']' returned non-zero exit status 1.

The above exception was the direct cause of the following exception:

Traceback (most recent call last):

File "/home/rootroot/.local/lib/python3.8/site-packages/whisper/transcribe.py", line 478, in cli

result = transcribe(model, audio_path, temperature=temperature, **args)

File "/home/rootroot/.local/lib/python3.8/site-packages/whisper/transcribe.py", line 122, in transcribe

mel = log_mel_spectrogram(audio, model.dims.n_mels, padding=N_SAMPLES)

File "/home/rootroot/.local/lib/python3.8/site-packages/whisper/audio.py", line 140, in log_mel_spectrogram

audio = load_audio(audio)

File "/home/rootroot/.local/lib/python3.8/site-packages/whisper/audio.py", line 60, in load_audio

raise RuntimeError(f"Failed to load audio: {e.stderr.decode()}") from e

built with gcc 9 (Ubuntu 9.4.0-1ubuntu1~20.04.1)

libavutil 56. 31.100 / 56. 31.100

libavcodec 58. 54.100 / 58. 54.100

libavformat 58. 29.100 / 58. 29.100

libavdevice 58. 8.100 / 58. 8.100

libavfilter 7. 57.100 / 7. 57.100

libavresample 4. 0. 0 / 4. 0. 0

libswscale 5. 5.100 / 5. 5.100

libswresample 3. 5.100 / 3. 5.100

libpostproc 55. 5.100 / 55. 5.100

audio.mp3: No such file or directory

built with gcc 9 (Ubuntu 9.4.0-1ubuntu1~20.04.1)

libavutil 56. 31.100 / 56. 31.100

libavcodec 58. 54.100 / 58. 54.100

libavformat 58. 29.100 / 58. 29.100

libavdevice 58. 8.100 / 58. 8.100

libavfilter 7. 57.100 / 7. 57.100

libavresample 4. 0. 0 / 4. 0. 0

libswscale 5. 5.100 / 5. 5.100

libswresample 3. 5.100 / 3. 5.100

libpostproc 55. 5.100 / 55. 5.100

audio.mp3: No such file or directory

rootroot@rootroot-X99-Turbo:~$

rootroot@rootroot-X99-Turbo:~$ whisper chi.mp4 --model medium --language Chinese

Traceback (most recent call last):

File "/home/rootroot/.local/lib/python3.8/site-packages/whisper/audio.py", line 58, in load_audio

out = run(cmd, capture_output=True, check=True).stdout

File "/usr/lib/python3.8/subprocess.py", line 516, in run

raise CalledProcessError(retcode, process.args,

subprocess.CalledProcessError: Command '['ffmpeg', '-nostdin', '-threads', '0', '-i', 'chi.mp4', '-f', 's16le', '-ac', '1', '-acodec', 'pcm_s16le', '-ar', '16000', '-']' returned non-zero exit status 1.

The above exception was the direct cause of the following exception:

Traceback (most recent call last):

File "/home/rootroot/.local/lib/python3.8/site-packages/whisper/transcribe.py", line 478, in cli

result = transcribe(model, audio_path, temperature=temperature, **args)

File "/home/rootroot/.local/lib/python3.8/site-packages/whisper/transcribe.py", line 122, in transcribe

mel = log_mel_spectrogram(audio, model.dims.n_mels, padding=N_SAMPLES)

File "/home/rootroot/.local/lib/python3.8/site-packages/whisper/audio.py", line 140, in log_mel_spectrogram

audio = load_audio(audio)

File "/home/rootroot/.local/lib/python3.8/site-packages/whisper/audio.py", line 60, in load_audio

raise RuntimeError(f"Failed to load audio: {e.stderr.decode()}") from e

built with gcc 9 (Ubuntu 9.4.0-1ubuntu1~20.04.1)

libavutil 56. 31.100 / 56. 31.100

libavcodec 58. 54.100 / 58. 54.100

libavformat 58. 29.100 / 58. 29.100

libavdevice 58. 8.100 / 58. 8.100

libavfilter 7. 57.100 / 7. 57.100

libavresample 4. 0. 0 / 4. 0. 0

libswscale 5. 5.100 / 5. 5.100

libswresample 3. 5.100 / 3. 5.100

libpostproc 55. 5.100 / 55. 5.100

chi.mp4: No such file or directory

built with gcc 9 (Ubuntu 9.4.0-1ubuntu1~20.04.1)

libavutil 56. 31.100 / 56. 31.100

libavcodec 58. 54.100 / 58. 54.100

libavformat 58. 29.100 / 58. 29.100

libavdevice 58. 8.100 / 58. 8.100

libavfilter 7. 57.100 / 7. 57.100

libavresample 4. 0. 0 / 4. 0. 0

libswscale 5. 5.100 / 5. 5.100

libswresample 3. 5.100 / 3. 5.100

libpostproc 55. 5.100 / 55. 5.100

chi.mp4: No such file or directory

rootroot@rootroot-X99-Turbo:~$

rootroot@rootroot-X99-Turbo:~$ sudo apt-get install ffmpeg

Reading package lists... Done

Building dependency tree

Reading state information... Done

ffmpeg is already the newest version (7:4.2.7-0ubuntu0.1).

0 upgraded, 0 newly installed, 0 to remove and 30 not upgraded.

rootroot@rootroot-X99-Turbo:~$

rootroot@rootroot-X99-Turbo:~$ ll *.mp4

-rwx------ 1 rootroot rootroot 3465644 1月 12 01:28 chs.mp4*

rootroot@rootroot-X99-Turbo:~$

rootroot@rootroot-X99-Turbo:~$ whisper chs.mp4 --model medium --language Chinese

[00:00.000 --> 00:01.400] 前段時間有個巨石鴻吼

[00:01.400 --> 00:03.000] 某某是男人最好的衣妹

[00:03.000 --> 00:04.800] 這裡的某某可以替換為減肥

[00:04.800 --> 00:07.800] 長髮西裝考研術唱永潔無間等等等等

[00:07.800 --> 00:09.200] 我聽到最新的一個說法是

[00:09.200 --> 00:12.000] 微分碎蓋加口罩加半框眼鏡加春風衣

[00:12.000 --> 00:13.400] 等於男人最好的衣妹

[00:13.400 --> 00:14.400] 大概也就前幾年

[00:14.400 --> 00:17.400] 春風衣還和格子襯衫並列為程序員穿搭精華

[00:17.400 --> 00:20.000] 紫紅色春風衣還被譽為廣場舞大媽標配

[00:20.000 --> 00:21.600] 路透牌還是我爹這個年紀的人

[00:21.600 --> 00:22.800] 才會願意買的牌子

[00:22.800 --> 00:24.400] 不知道風向為啥變得這麼快

[00:24.400 --> 00:26.800] 為啥這東西突然變成男生逆襲神器

[00:26.800 --> 00:27.800] 時尚潮流單品

[00:27.800 --> 00:29.400] 後來我翻了一下小紅書就懂了

[00:29.400 --> 00:30.400] 時尚這個時期

[00:30.400 --> 00:31.600] 重點不在於衣服

[00:31.600 --> 00:32.200] 在於人

[00:32.200 --> 00:34.600] 先在小紅書上面和春風衣相關的筆記

[00:34.600 --> 00:36.200] 照片裡的男生都是這樣的

[00:36.200 --> 00:37.000] 這樣的

[00:37.000 --> 00:38.000] 還有這樣的

[00:38.000 --> 00:39.400] 你們哪裡是看穿搭的

[00:39.400 --> 00:40.600] 你們明明是看臉

[00:40.600 --> 00:41.800] 就這個造型這個年齡

[00:41.800 --> 00:44.000] 你換上老頭衫也能穿出氛圍感好嗎

[00:44.000 --> 00:46.600] 我又想起了當年郭德綱老師穿季凡西的殘劇

[00:46.600 --> 00:48.600] 這個世界對我們這些長得不好看的人

[00:48.600 --> 00:49.600] 還真是苛刻的

[00:49.600 --> 00:52.000] 所以說我總結了一下春風衣傳達的要領

[00:52.200 --> 00:54.400] 大概就是一張白鏡且人畜無憾的臉

[00:54.400 --> 00:55.200] 充足的髮量

[00:55.200 --> 00:56.200] 纖細的體型

[00:56.200 --> 00:58.200] 當然身上的春風衣還得是駱駝的

[00:58.200 --> 00:59.400] 去年在戶外用品界

[00:59.400 --> 01:00.200] 最頂流的

[01:00.200 --> 01:01.200] 既不是鳥橡樹

[01:01.200 --> 01:02.800] 也不是有校服之稱的北面

[01:02.800 --> 01:04.200] 或者老臺頂流哥倫比亞

[01:04.200 --> 01:05.000] 而是駱駝

[01:05.000 --> 01:07.200] 雙11 駱駝在天貓戶外服飾品類

[01:07.200 --> 01:09.000] 拿下銷售額和銷量雙料冠軍

[01:09.000 --> 01:10.200] 銷量達到百萬幾

[01:10.200 --> 01:10.800] 再抖音

[01:10.800 --> 01:13.400] 駱駝銷售同比增幅高達296%

[01:13.400 --> 01:16.200] 旗下主打的三合一高性價比春風衣成為爆品

[01:22.600 --> 01:23.200] 至於線下

[01:23.200 --> 01:24.400] 還是網友總覺得好

[01:24.400 --> 01:26.800] 如今在南方街頭的駱駝比沙漠裡的都多

[01:30.000 --> 01:31.200] 至於駱駝為啥這麼火

[01:31.200 --> 01:32.000] 便宜啊

[01:32.000 --> 01:33.600] 拿賣得最好的丁珍同款

[01:33.600 --> 01:35.600] 幻影黑三合一春風衣舉個例子

[01:35.600 --> 01:36.000] 線下買

[01:36.000 --> 01:37.600] 標牌價格2198

[01:37.600 --> 01:39.200] 但是跑到網上看一下

[01:39.200 --> 01:40.800] 標價就變成了699

[01:40.800 --> 01:41.400] 至於折扣

[01:41.400 --> 01:42.400] 日常也都是有的

[01:42.400 --> 01:43.600] 400出頭就能買到

[01:43.600 --> 01:45.200] 甚至有時候能递到300價

[01:45.200 --> 01:46.200] 要是你還嫌貴

[01:46.200 --> 01:48.400] 駱駝還有200塊出頭的單層春風衣

[01:48.400 --> 01:49.200] 就這個價格

[01:49.200 --> 01:51.800] 哥上海恐怕還不夠兩次City Walk的報名費

[01:51.800 --> 01:52.600] 看來這個價格

[01:52.600 --> 01:54.800] 再對比一下北面1000塊錢起步

[01:54.800 --> 01:56.000] 你就能理解為啥北面

[01:56.000 --> 01:58.200] 這麼快就被大學生踢出了校服序列了

[01:58.200 --> 02:00.400] 我不知道現在大學生每個月生活費多少

[02:00.400 --> 02:02.200] 反正按照我上學時候的生活費

[02:02.200 --> 02:03.200] 一個月不吃不喝

[02:03.200 --> 02:05.000] 也就買得起倆袖子加一個帽子

[02:05.000 --> 02:06.400] 難怪當年全是假北面

[02:06.400 --> 02:07.400] 現在都是真駱駝

[02:07.400 --> 02:08.800] 至少人家是正品啊

[02:08.800 --> 02:10.000] 我翻了一下社交媒體

[02:10.000 --> 02:11.200] 發現對駱駝的吐槽

[02:11.200 --> 02:12.000] 和買了駱駝的

[02:12.000 --> 02:13.400] 基本上是1比1的比例

[02:13.400 --> 02:15.000] 吐槽最多的就是衣服會掉色

[02:15.000 --> 02:15.800] 還會串色

[02:15.800 --> 02:17.000] 比如圖層洗個幾次

[02:17.000 --> 02:18.200] 穿個兩天就掉光了

[02:18.200 --> 02:19.600] 比如不同倉庫發的貨

[02:19.600 --> 02:20.600] 質量參差不齊

[02:20.600 --> 02:21.600] 買衣服還得看戶口

[02:21.600 --> 02:22.400] 聽出聲

[02:22.400 --> 02:23.600] 至於什麼做工比較差

[02:23.600 --> 02:24.800] 內膽多走線操

[02:24.800 --> 02:26.400] 不防水之類的就更多了

[02:26.400 --> 02:27.400] 但是這些吐槽

[02:27.400 --> 02:29.200] 並不意味著會影響駱駝的銷量

[02:29.200 --> 02:30.800] 甚至還會有不少自來水表示

[02:30.800 --> 02:32.600] 就這價格要啥子行車啊

[02:32.600 --> 02:34.000] 所謂性價比性價比

[02:34.000 --> 02:35.200] 脫離價位談性能

[02:35.200 --> 02:37.000] 這就不符合消費者的需求嘛

[02:37.000 --> 02:38.400] 無數次價格戰告訴我們

[02:38.400 --> 02:39.400] 只要肯降價

[02:39.400 --> 02:41.000] 就沒有賣不出去的產品

[02:41.000 --> 02:42.400] 一件衝鋒衣1000多

[02:42.400 --> 02:43.600] 你覺得平平無奇

[02:43.600 --> 02:45.000] 500多你覺得差點意思

[02:45.000 --> 02:46.400] 200塊你就秒下單了

[02:46.400 --> 02:47.000] 到99

[02:47.000 --> 02:48.400] 恐怕就要拼點手速了

[02:48.400 --> 02:49.600] 像衝鋒衣這個品類

[02:49.600 --> 02:50.800] 本來價格跨度就大

[02:50.800 --> 02:52.800] 北面最便宜的GORTEX衝鋒衣

[02:52.800 --> 02:53.800] 價格3000起步

[02:53.800 --> 02:55.200] 大概是同品牌最便宜

[02:55.200 --> 02:56.200] 衝鋒衣的三倍價格

[02:56.200 --> 02:57.200] 至於十足那樣

[02:57.200 --> 02:59.000] 搭載了GORTEX的硬殼起步價

[02:59.000 --> 03:00.000] 就要到4500

[03:00.000 --> 03:01.200] 而且同樣是GORTEX

[03:01.200 --> 03:02.800] 內部也有不同的系列和檔次

[03:02.800 --> 03:03.600] 做成衣服

[03:03.600 --> 03:05.600] 中間的差價恐怕就夠買兩件駱駝了

[03:05.600 --> 03:06.600] 至於智能控溫

[03:06.600 --> 03:07.400] 防水拉鍊

[03:07.400 --> 03:08.000] 全壓膠

[03:08.000 --> 03:09.800] 更加不可能出現在駱駝這裡了

[03:09.800 --> 03:11.800] 至少不會是300 400的駱駝身上會有的

[03:11.800 --> 03:12.800] 有的價外的衣服

[03:12.800 --> 03:14.200] 買的就是一個放棄幻想

[03:14.200 --> 03:15.800] 吃到肚子裡的科技魚很活

[03:15.800 --> 03:17.000] 是能給你省錢的

[03:17.000 --> 03:18.400] 穿在身上的科技魚很活

[03:18.400 --> 03:20.000] 裝裝件件都是要加錢的

[03:20.000 --> 03:21.600] 所以正如羅曼羅蘭所說

[03:21.600 --> 03:23.200] 這世界上只有一種英雄主義

[03:23.200 --> 03:24.800] 就是在認清了駱駝的本質以後

[03:24.800 --> 03:26.000] 依然選擇買駱駝

[03:26.000 --> 03:27.000] 關於駱駝的火爆

[03:27.000 --> 03:28.200] 我有一些小小的看法

[03:28.200 --> 03:29.000] 駱駝這個東西

[03:29.000 --> 03:30.400] 它其實就是個潮牌

[03:30.400 --> 03:32.000] 看看它的營銷方式就知道了

[03:32.000 --> 03:33.000] 現在打開小黃書

[03:33.000 --> 03:35.000] 日常可以看到駱駝穿搭是這樣的

[03:35.000 --> 03:36.600] 加一點氛圍感是這樣的

[03:36.600 --> 03:37.400] 對比一下

[03:37.400 --> 03:39.000] 其他品牌的風格是這樣的

[03:39.000 --> 03:39.800] 這樣的

[03:39.800 --> 03:41.200] 其實對比一下就知道了

[03:41.200 --> 03:42.600] 其他品牌突出一個時程

[03:42.600 --> 03:44.200] 能防風就一定要講防風

[03:44.200 --> 03:46.000] 能扛動就一定要講扛動

[03:46.000 --> 03:47.400] 但駱駝在營銷的時候

[03:47.400 --> 03:49.200] 主打的就是一個城市戶外風

[03:49.200 --> 03:50.400] 雖然造型是春風衣

[03:50.400 --> 03:52.200] 但場景往往是在城市裡

[03:52.200 --> 03:54.200] 哪怕在野外也要突出一個風和日麗

[03:54.200 --> 03:55.000] 陽光美媚

[03:55.000 --> 03:56.400] 至少不會在明顯的嚴寒

[03:56.400 --> 03:58.000] 高海拔或是惡劣氣候下

[03:58.200 --> 04:00.200] 如果用一個詞形容駱駝的營銷風格

[04:00.200 --> 04:01.000] 那就是清洗

[04:01.000 --> 04:03.000] 或者說他很理解自己的消費者是誰

[04:03.000 --> 04:04.000] 需要什麼產品

[04:04.000 --> 04:05.200] 從使用場景來說

[04:05.200 --> 04:06.600] 駱駝的消費者買春風衣

[04:06.600 --> 04:08.800] 不是真的有什麼大風大雨要去應對

[04:08.800 --> 04:11.000] 春風衣的作用是下雨沒帶傘的時候

[04:11.000 --> 04:12.000] 臨時頂個幾分鐘

[04:12.000 --> 04:13.600] 讓你能圖書館跑回宿舍

[04:13.600 --> 04:15.000] 或者是冬天騎電動車

[04:15.000 --> 04:16.200] 被風吹得不行的時候

[04:16.200 --> 04:17.200] 稍微扛一下風

[04:17.200 --> 04:18.400] 不至於體感太冷

[04:18.400 --> 04:19.800] 當然他們也會出門

[04:19.800 --> 04:21.800] 但大部分時候也都是去別的城市

[04:21.800 --> 04:24.000] 或者在城市周邊搞搞簡單的徒步

[04:24.000 --> 04:26.000] 這種情況下穿個駱駝已經夠了

[04:26.000 --> 04:27.200] 從購買動機來說

[04:27.200 --> 04:29.200] 駱駝就更沒有必要上那些應回科技了

[04:29.200 --> 04:31.000] 消費者買駱駝買的是個什麼呢

[04:31.000 --> 04:32.200] 不是春風衣的功能性

[04:32.200 --> 04:33.400] 而是春風衣的造型

[04:33.400 --> 04:34.400] 寬鬆的版型

[04:34.400 --> 04:36.400] 能精準遮住微微隆起的小肚子

[04:36.400 --> 04:37.400] 棱角分明的質感

[04:37.400 --> 04:39.400] 能隱藏一切不完美的身體線條

[04:39.400 --> 04:41.400] 顯瘦的副作用就是顯年輕

[04:41.400 --> 04:42.600] 再配上一條牛仔褲

[04:42.600 --> 04:43.800] 配上一雙大黃靴

[04:43.800 --> 04:45.200] 大學生的氣質就出來了

[04:45.200 --> 04:46.200] 要是自拍的時候

[04:46.200 --> 04:47.800] 再配上大學宿舍洗素臺

[04:47.800 --> 04:49.200] 那永遠擦不乾淨的鏡子

[04:49.200 --> 04:50.600] 瞬間青春無敵了

[04:50.800 --> 04:51.800] 說的更直白一點

[04:51.800 --> 04:53.200] 人家買的是個簡靈神器

[04:53.200 --> 04:53.800] 所以說

[04:53.800 --> 04:56.000] 吐槽穿駱駝都是假戶外愛好者的人

[04:56.000 --> 04:57.600] 其實並沒有理解駱駝的定位

[04:57.600 --> 04:59.800] 駱駝其實是給了想要入門山系穿搭

[04:59.800 --> 05:01.800] 想要追逐流行的人一個最平價

[05:01.800 --> 05:03.000] 決策成本最低的選擇

[05:03.000 --> 05:04.800] 至於那些真正的硬核戶外愛好者

[05:04.800 --> 05:05.800] 駱駝既沒有能力

[05:05.800 --> 05:07.200] 也沒有打算觸打他們

[05:07.200 --> 05:08.000] 反過來說

[05:08.000 --> 05:09.600] 那些自駕穿越邊疆國道

[05:09.600 --> 05:11.800] 或者去奧爾卑斯山區登山探險的人

[05:11.800 --> 05:13.600] 也不太可能在戶外服飾上省錢

[05:13.600 --> 05:15.000] 畢竟光是交通住宿

[05:15.400 --> 05:16.400] 成本就不低了

[05:16.400 --> 05:17.200] 對他們來說

[05:17.200 --> 05:19.000] 戶外裝備很多時候是保命用的

[05:19.000 --> 05:21.000] 也就不存在跟風奧造型的必要了

[05:21.000 --> 05:22.200] 最後我再說個題外話

[05:22.200 --> 05:24.200] 年輕人追捧駱駝一個隱藏的原因

[05:24.200 --> 05:25.800] 其實是羽絨服越來越貴了

[05:25.800 --> 05:26.600] 有媒體統計

[05:26.600 --> 05:30.000] 現在國產羽絨服的平均售價已經高達881元

[05:30.000 --> 05:32.000] 波斯登均價最高接近2000元

[05:32.000 --> 05:32.800] 而且過去幾年

[05:32.800 --> 05:34.800] 國產羽絨服品牌都在轉向高端化

[05:34.800 --> 05:37.000] 羽絨服市場分為8000元以上的奢侈級

[05:37.000 --> 05:38.400] 2000元以下的大眾級

[05:38.400 --> 05:39.800] 而在中間的高端級

[05:39.800 --> 05:41.200] 國產品牌一直沒有存在感

[05:41.200 --> 05:42.200] 所以過去幾年

[05:42.200 --> 05:43.600] 波斯登天工人這些品牌

[05:43.600 --> 05:45.200] 都把2000元到8000元這個市場

[05:45.200 --> 05:46.600] 當成未來的發展趨勢

[05:46.600 --> 05:48.000] 東新證券研報顯示

[05:48.000 --> 05:49.600] 從2018到2021年

[05:49.600 --> 05:52.200] 波斯登均價4年漲幅達到60%以上

[05:52.200 --> 05:53.200] 過去5個菜年

[05:53.200 --> 05:55.000] 這個品牌的營銷開支從20多億

[05:55.000 --> 05:56.000] 漲到了60多億

[05:56.000 --> 05:57.200] 羽絨服價格往上走

[05:57.200 --> 05:59.200] 年輕消費者就開始拋棄羽絨服

[05:59.200 --> 06:00.400] 購買平價衝鋒衣

[06:00.400 --> 06:02.200] 裡面再穿個普通價外的瑤麗絨

[06:02.200 --> 06:03.400] 或者羽絨小夾克

[06:03.400 --> 06:05.200] 也不比大幾千的羽絨服差多少

[06:05.200 --> 06:05.800] 說到底

[06:05.800 --> 06:07.000] 現在消費社會發達了

[06:07.000 --> 06:08.000] 沒有什麼需求是

[06:08.000 --> 06:09.600] 一定要某種特定的解決方案

[06:09.600 --> 06:11.600] 特定價位的商品才能實現的

[06:11.600 --> 06:12.200] 要保暖

[06:12.200 --> 06:13.200] 羽絨服固然很好

[06:13.200 --> 06:15.200] 但衝鋒衣加一些內搭也很暖和

[06:15.200 --> 06:16.000] 要時尚

[06:16.000 --> 06:18.000] 大幾千塊錢的設計師品牌非常不錯

[06:18.000 --> 06:19.400] 但350的拼多多服飾

[06:19.400 --> 06:20.600] 搭得好也能出彩

[06:20.600 --> 06:21.600] 要去野外徒步

[06:21.600 --> 06:23.000] 花五六千買鳥也可以

[06:23.000 --> 06:25.200] 但迪卡農也足以應付大多數狀況

[06:25.200 --> 06:25.800] 所以說

[06:25.800 --> 06:27.600] 花高價買衝鋒衣當然也OK

[06:27.600 --> 06:28.600] 三四百買件駱駝

[06:28.600 --> 06:29.800] 也是可以接受的選擇

[06:29.800 --> 06:32.000] 何況駱駝也多多少少有一些功能性

[06:32.000 --> 06:33.800] 畢竟它再怎麼樣還是個衝鋒衣

[06:33.800 --> 06:34.800] 理解了這個事情

[06:34.800 --> 06:36.800] 就很容易分辨什麼是智商稅的

[06:36.800 --> 06:38.800] 那些向你灌輸非某個品牌不用

[06:38.800 --> 06:39.800] 告訴你某個需求

[06:39.800 --> 06:41.400] 只有某個產品才能滿足

[06:41.400 --> 06:42.200] 某個品牌

[06:42.200 --> 06:44.400] 就是某個品牌絕對的比試鏈頂端

[06:44.400 --> 06:46.800] 這類銀銷的智商稅含量必然是很高的

[06:46.800 --> 06:48.800] 它的目的是剝奪你選擇的權利

[06:48.800 --> 06:51.200] 讓你主動放棄比價和尋找平梯的想法

[06:51.200 --> 06:53.000] 從而避免與其他品牌競爭

[06:53.000 --> 06:54.200] 而沒有競爭的市場

[06:54.200 --> 06:56.200] 才是智商稅含量最高的市場

[06:56.200 --> 06:57.400] 消費商業洞穴

[06:57.400 --> 06:58.400] 禁在IC實驗室

[06:58.400 --> 06:59.000] 我是館長

[06:59.000 --> 07:00.000] 我們下期再見

rootroot@rootroot-X99-Turbo:~$

rootroot@rootroot-X99-Turbo:~$ time(whisper chs.mp4 --model medium --language Chinese)

https://www.toutiao.com/article/7189209812264075835/?app=news_article\&timestamp=1706203570\&use_new_style=1\&req_id=20240126012609901ACEF7F5666533AA21\&group_id=7189209812264075835\&tt_from=mobile_qq\&utm_source=mobile_qq\&utm_medium=toutiao_android\&utm_campaign=client_share\&share_token=5e0cda89-00c5-40fe-afa0-c3c88dd056c4\&source=m_redirect

已达到人类水准语音识别模型的whisper，真的有这么厉害吗？

transcribe函数的language目前支持99种语言，如下：

"en": "english","zh": "chinese",

"de": "german","es": "spanish",

"ru": "russian","ko": "korean",

"fr": "french","ja": "japanese",

"pt": "portuguese","tr": "turkish",

"pl": "polish","ca": "catalan",

"nl": "dutch","ar": "arabic",

"sv": "swedish","it": "italian",

"id": "indonesian","hi": "hindi",

"fi": "finnish","vi": "vietnamese",

"he": "hebrew","uk": "ukrainian",

"el": "greek","ms": "malay",

"cs": "czech","ro": "romanian",

"da": "danish","hu": "hungarian",

"ta": "tamil","no": "norwegian",

"th": "thai","ur": "urdu",

"hr": "croatian","bg": "bulgarian",

"lt": "lithuanian","la": "latin",

"mi": "maori","ml": "malayalam",

"cy": "welsh","sk": "slovak",

"te": "telugu","fa": "persian",

"lv": "latvian","bn": "bengali",

"sr": "serbian","az": "azerbaijani",

"sl": "slovenian","kn": "kannada",

"et": "estonian","mk": "macedonian",

"br": "breton","eu": "basque",

"is": "icelandic","hy": "armenian",

"ne": "nepali","mn": "mongolian",

"bs": "bosnian","kk": "kazakh",

"sq": "albanian","sw": "swahili",

"gl": "galician","mr": "marathi",

"pa": "punjabi","si": "sinhala",

"km": "khmer","sn": "shona",

"yo": "yoruba","so": "somali",

"af": "afrikaans","oc": "occitan",

"ka": "georgian","be": "belarusian",

"tg": "tajik","sd": "sindhi",

"gu": "gujarati","am": "amharic",

"yi": "yiddish","lo": "lao",

"uz": "uzbek","fo": "faroese",

"ht": "haitian creole","ps": "pashto",

"tk": "turkmen","nn": "nynorsk",

"mt": "maltese","sa": "sanskrit",

"lb": "luxembourgish","my": "myanmar",

"bo": "tibetan","tl": "tagalog",

"mg": "malagasy","as": "assamese",

"tt": "tatar","haw": "hawaiian",

"ln": "lingala","ha": "hausa",

"ba": "bashkir","jw": "javanese","su": "sundanese",

官方还提供了另外一种调用方案：

import whisper

model = whisper.load_model("base")

load audio and pad/trim it to fit 30 seconds

audio = whisper.load_audio("audio.mp3")

audio = whisper.pad_or_trim(audio)

make log-Mel spectrogram and move to the same device as the model

mel = whisper.log_mel_spectrogram(audio).to(model.device)

detect the spoken language

_, probs = model.detect_language(mel)

print(f"Detected language: {max(probs, key=probs.get)}")

decode the audio

options = whisper.DecodingOptions(language='Chinese')

result = whisper.decode(model, mel, options)

print the recognized text

print(result.text)

参考资料：

https://www.toutiao.com/article/7229151806801248807/?app=news_article\&timestamp=1706203733\&use_new_style=1\&req_id=20240126012853D9D3D4539BEF1333DBCC\&group_id=7229151806801248807\&tt_from=mobile_qq\&utm_source=mobile_qq\&utm_medium=toutiao_android\&utm_campaign=client_share\&share_token=085ce76c-b23a-4609-b2d0-d18c8d7ab8f8\&source=m_redirect

C++版本人工智能实时语音转文字(字幕/语音识别)Whisper.cpp实践

【WINDOWS，大模型需要10GB】

https://blog.csdn.net/hhy321/article/details/134897967?spm=1001.2101.3001.6650.2\&utm_medium=distribute.wap_relevant.none-task-blog-2\~default\~CTRLIST\~Rate-2-134897967-blog-130001848.237^v3^wap_relevant_t0_download\&depth_1-utm_source=distribute.wap_relevant.none-task-blog-2\~default\~CTRLIST\~Rate-2-134897967-blog-130001848.237^v3^wap_relevant_t0_download\&share_token=845e69c5-c625-4834-8faa-08f1f29f55b2

【小沐学Python】Python实现语音识别（Whisper）

https://blog.csdn.net/xkukeer/article/details/130227944?share_token=f48bfb40-9399-4375-894e-3ecf96d1c51d

openai的whisper语音识别介绍

第三步，选择使用的模型。

官方说有5种模型，其中4种是English-only模型，但是实测english-only也可以支持中文（只测了base可以支持中文，其他的没测但应该也可以）

虽说支持中文，但是也有不理想的地方，中文的识别错误率（WER (Word Error Rate)）还不低，在所有支持语言的大概排中游水平。

第四步，具体使用

有好几种方法：

1、命令行模式

whisper audio.flac audio.mp3 audio.wav --model medium

对于非英文语言，加上--language参数，例如日语

whisper japanese.wav --language Japanese

支持的语言类型还挺多的

【WINDOWS】

https://blog.csdn.net/liaoqingjian/article/details/132474687?share_token=e6ad6f74-2fab-45c5-bdb5-40b48fe2cd79

whisper 语音识别项目部署

https://www.toutiao.com/article/7327918175801164325/?app=news_article\&timestamp=1706203446\&use_new_style=1\&req_id=202401260124058D2D3B0452AC9B3435B3\&group_id=7327918175801164325\&tt_from=mobile_qq\&utm_source=mobile_qq\&utm_medium=toutiao_android\&utm_campaign=client_share\&share_token=ad4cdc74-1590-4a7b-b020-14f9186f9ef2\&source=m_redirect

Whisper对于中文语音识别与转写中文文本优化的实践(Python3.10)

【WINDOWS】

https://www.toutiao.com/article/7276749520275456572/?app=news_article\&timestamp=1706203504\&use_new_style=1\&req_id=2024012601250342BCD0F3D434AA335380\&group_id=7276749520275456572\&tt_from=mobile_qq\&utm_source=mobile_qq\&utm_medium=toutiao_android\&utm_campaign=client_share\&share_token=5bc13cbe-db1d-4883-bff4-b01f258dd1c2\&source=m_redirect

语音转文字软件Whisper，实时自动语音识别，音频视频文案提取

20240127在ubuntu20.04.6下配置whisper

load audio and pad/trim it to fit 30 seconds

make log-Mel spectrogram and move to the same device as the model

detect the spoken language

decode the audio

print the recognized text