https://github.com/modelscope/FunASR/blob/main/README_zh.md
https://github.com/modelscope/FunASR/blob/main/model_zoo/readme_zh.md
PyTorch / 2.3.0 / 3.12(ubuntu22.04) / 12.1
Paraformer分角色语音识别-中文-通用
https://www.modelscope.cn/models/iic/speech_paraformer-large-vad-punc-spk_asr_nat-zh-cn
安装ffmpeg
bash
source /etc/network_turbo
conda install x264 ffmpeg -c conda-forge -y
# 或者
conda install -c https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge/ x264 ffmpeg -y
bash
source /etc/network_turbo
pip install torchaudio
pip install -U funasr
python -c "import torchaudio; print(torchaudio.__version__)"
python -c "import funasr; print(funasr.__version__)"1 模型下载
模型下载：https://modelscope.cn/models/iic/speech_paraformer-large-vad-punc-spk_asr_nat-zh-cn/files
使用SDK下载下载：
开始前安装
bash
source /etc/network_turbo
pip install modelscope脚本下载
python
# source /etc/network_turbo
from modelscope import snapshot_download
# 指定模型的下载路径
cache_dir = '/root/autodl-tmp'
# 调用 snapshot_download 函数下载模型
model_dir = snapshot_download('iic/speech_paraformer-large-vad-punc-spk_asr_nat-zh-cn', cache_dir=cache_dir)
print(f"模型已下载到: {model_dir}")2 音频识别测试
音频下载
bash
wget https://isv-data.oss-cn-hangzhou.aliyuncs.com/ics/MaaS/ASR/test_audio/vad_example.wav
python
from funasr import AutoModel
# paraformer-zh is a multi-functional asr model
# use vad, punc, spk or not as you need
# model = AutoModel(model="paraformer-zh", model_revision="v2.0.4",
model = AutoModel(model="/root/autodl-tmp/iic/speech_paraformer-large-vad-punc-spk_asr_nat-zh-cn", model_revision="v2.0.4",
vad_model="fsmn-vad", vad_model_revision="v2.0.4",
punc_model="ct-punc-c", punc_model_revision="v2.0.4",
# spk_model="cam++", spk_model_revision="v2.0.2",
)
# res = model.generate(input=f"{model.model_path}/example/asr_example.wav",
res = model.generate(input=f"vad_example.wav",
batch_size_s=300,
hotword='魔搭')
print(res)结果如下：
{'key': 'vad_example', 'text': '试错的过程很简单啊，今特别是今天冒名插血卡的同学，你们可以听到后面的有专门的活动课，它会大大降低你的思错成本。其实你也可以不要来听课，为什么你自己写嘛？我先今天写五个点，我就实试实验一下，反正这五个点不行，我再写五个点，再是再不行，那再写五个点嘛。你总会所谓的活动大神和所谓的高手都是只有一个，把所有的错。所有的坑全部趟一遍，留下正确的你就是所谓的大神明白吗？所以说关于活动通过这一块，我只送给你们四个字啊，换位思考。如果说你要想降低你的试错成本，今天来这里你们就是对的。因为有创企创需要搞这个机会。所以说关于活动过于不过这个问题或者活动很难通过这个话题。呃，如果真的要坐下来聊的话，要聊一天。但是我觉得我刚才说的四个字足够好，谢谢。好，非常感谢那个三毛老师的回答啊，三毛老师说我们在整个店铺的这个活动当中，我们要学会换位思考。其实。', 'timestamp': \[\[380, 620\], \[640, 740\], \[740, 940\], \[940, 1020\], \[1020, 1260\], \[1500, 1740\], \[1740, 1840\], \[1840, 2135\], \[2830, 3010\], \[3010, 3210\], \[3210, 3290\], \[3290, 3370\], \[3370, 3470\], \[3470, 3590\], \[3590, 3830\], \[3950, 4130\], \[4130, 4270\], \[4270, 4350\], \[4350, 4470\], \[4470, 4590\], \[4590, 4690\], \[4690, 4770\], \[4770, 5010\], \[5250, 5410\], \[5410, 5530\], \[5530, 5650\], \[5650, 5975\], \[6670, 6830\], \[6830, 6970\], \[6970, 7110\], \[7110, 7230\], \[7230, 7470\], \[7490, 7730\], \[8070, 8310\], \[8310, 8430\], \[8430, 8670\], \[8690, 8910\], \[8910, 9030\], \[9030, 9270\], \[9550, 9750\], \[9750, 9910\], \[9910, 10110\], \[10110, 10350\], \[10670, 10910\], \[10950, 11130\], \[11130, 11250\], \[11250, 11370\], \[11370, 11490\], \[11490, 11630\], \[11630, 11730\], \[11730, 11970\], \[12310, 12490\], \[12490, 12610\], \[12610, 12710\], \[12710, 12790\], \[12790, 12910\], \[12910, 13110\], \[13110, 13270\], \[13270, 13350\], \[13350, 13490\], \[13490, 13630\], \[13630, 13870\], \[14030, 14250\], \[14250, 14350\], \[14350, 14589\], \[14630, 14850\], \[14850, 14950\], \[14950, 15070\], \[15070, 15250\], \[15250, 15490\], \[15950, 16150\], \[16150, 16390\], \[16390, 16490\], \[16490, 16610\], \[16610, 16750\], \[16750, 16850\], \[16850, 16970\], \[16970, 17210\], \[17270, 17390\], \[17390, 17570\], \[17570, 17810\], \[17990, 18230\], \[18310, 18410\], \[18410, 18550\], \[18550, 18650\], \[18650, 18870\], \[18870, 19010\], \[19010, 19090\], \[19090, 19190\], \[19190, 19310\], \[19310, 19390\], \[19390, 19490\], \[19490, 19590\], \[19590, 19830\], \[19970, 20130\], \[20130, 20250\], \[20250, 20430\], \[20430, 20550\], \[20550, 20670\], \[20670, 20910\], \[21090, 21270\], \[21270, 21510\], \[21510, 21650\], \[21650, 21750\], \[21750, 21990\], \[22470, 22609\], \[22609, 22710\], \[22710, 22890\], \[22890, 22970\], \[22970, 23070\], \[23070, 23190\], \[23190, 23485\], \[24060, 24280\], \[24280, 24380\], \[24380, 24620\], \[25120, 25360\], \[25500, 25660\], \[25660, 25985\], \[27030, 27230\], \[27230, 27410\], \[27410, 27510\], \[27510, 27750\], \[27810, 27990\], \[27990, 28150\], \[28150, 28270\], \[28270, 28350\], \[28350, 28430\], \[28430, 28755\], \[30180, 30320\], \[30320, 30560\], \[30600, 30720\], \[30720, 30840\], \[30840, 30940\], \[30940, 31235\], \[32020, 32260\], \[32280, 32440\], \[32440, 32620\], \[32620, 32700\], \[32700, 32940\], \[33200, 33340\], \[33340, 33440\], \[33440, 33560\], \[33560, 33800\], \[33960, 34160\], \[34160, 34360\], \[34360, 34600\], \[34800, 35000\], \[35000, 35240\], \[35540, 35700\], \[35700, 35900\], \[35900, 36000\], \[36000, 36180\], \[36180, 36420\], \[36440, 36600\], \[36600, 36700\], \[36700, 36840\], \[36840, 36940\], \[36940, 37020\], \[37020, 37100\], \[37100, 37200\], \[37200, 37455\], \[38480, 38600\], \[38600, 38720\], \[38720, 38960\], \[39180, 39320\], \[39320, 39440\], \[39440, 39560\], \[39560, 39740\], \[39740, 39840\], \[39840, 40040\], \[40040, 40220\], \[40220, 40340\], \[40340, 40440\], \[40440, 40540\], \[40540, 40620\], \[40620, 40840\], \[40840, 40980\], \[40980, 41120\], \[41120, 41220\], \[41220, 41340\], \[41340, 41460\], \[41460, 41620\], \[41620, 41800\], \[41800, 41920\], \[41920, 42160\], \[42340, 42580\], \[42720, 42940\], \[42940, 43100\], \[43100, 43240\], \[43240, 43480\], \[43800, 44020\], \[44020, 44200\], \[44200, 44440\], \[44740, 44960\], \[44960, 45140\], \[45140, 45380\], \[45440, 45680\], \[45680, 45860\], \[45860, 45980\], \[45980, 46080\], \[46080, 46220\], \[46220, 46360\], \[46360, 46460\], \[46460, 46725\], \[47580, 47760\], \[47760, 48000\], \[48040, 48220\], \[48220, 48340\], \[48340, 48580\], \[48660, 48800\], \[48800, 48920\], \[48920, 49000\], \[49000, 49080\], \[49080, 49260\], \[49260, 49495\], \[50180, 50300\], \[50300, 50540\], \[50600, 50780\], \[50780, 50900\], \[50900, 51140\], \[51180, 51320\], \[51320, 51440\], \[51440, 51520\], \[51520, 51640\], \[51640, 51740\], \[51740, 51820\], \[51820, 51960\], \[51960, 52200\], \[52400, 52560\], \[52560, 52640\], \[52640, 52740\], \[52740, 52860\], \[52860, 52980\], \[52980, 53100\], \[53100, 53280\], \[53280, 53380\], \[53380, 53500\], \[53500, 53620\], \[53620, 53860\], \[54080, 54220\], \[54220, 54320\], \[54320, 54400\], \[54400, 54640\], \[54640, 54760\], \[54760, 54880\], \[54880, 54980\], \[54980, 55100\], \[55100, 55220\], \[55220, 55380\], \[55380, 55520\], \[55520, 55700\], \[55700, 55820\], \[55820, 55900\], \[55900, 56060\], \[56060, 56295\], \[57010, 57210\], \[57210, 57310\], \[57310, 57430\], \[57430, 57530\], \[57530, 57650\], \[57650, 57890\], \[57950, 58130\], \[58130, 58230\], \[58230, 58330\], \[58330, 58430\], \[58430, 58530\], \[58530, 58630\], \[58630, 58750\], \[58750, 58950\], \[58950, 59050\], \[59050, 59345\], \[60070, 60230\], \[60230, 60330\], \[60330, 60470\], \[60470, 60570\], \[60570, 60810\], \[60970, 61170\], \[61170, 61290\], \[61290, 61410\], \[61410, 61490\], \[61490, 61610\], \[61610, 61770\], \[61770, 61890\], \[61890, 62130\], \[62230, 62430\], \[62430, 62670\], \[63030, 63270\], \[63290, 63430\], \[63430, 63670\], \[63890, 64130\], \[64170, 64270\], \[64270, 64470\], \[64470, 64570\], \[64570, 64750\], \[64750, 64870\], \[64870, 65050\], \[65050, 65170\], \[65170, 65290\], \[65290, 65410\], \[65410, 65530\], \[65530, 65610\], \[65610, 65710\], \[65710, 65830\], \[65830, 66010\], \[66010, 66110\], \[66110, 66230\], \[66230, 66310\], \[66310, 66410\], \[66410, 66550\], \[66550, 66650\], \[66650, 66890\], \[67110, 67350\], \[67410, 67550\], \[67550, 67690\], \[67690, 67790\], \[67790, 67910\], \[67910, 68030\], \[68030, 68110\], \[68110, 68210\], \[68210, 68310\], \[68310, 68490\], \[68490, 68610\], \[68610, 68750\], \[68750, 68850\], \[68850, 68930\], \[68930, 69030\], \[69030, 69150\], \[69150, 69290\], \[69290, 69450\], \[69450, 69610\], \[69610, 69750\], \[69750, 69990\], \[69990, 70110\], \[70110, 70315\]\]}
FunASR安装
bash
source /etc/network_turbo
git clone https://github.com/alibaba/FunASR.git && cd FunASR进入到：FunASR/examples/industrial_data_pretraining/paraformer