Spaces:
Running
Running
Upload 2 files
Browse files- app.py +97 -91
- requirements.txt +6 -3
app.py
CHANGED
|
@@ -1,91 +1,97 @@
|
|
| 1 |
-
import gradio as gr
|
| 2 |
-
import numpy as np
|
| 3 |
-
import torch
|
| 4 |
-
import torchaudio
|
| 5 |
-
import re
|
| 6 |
-
from funasr import AutoModel
|
| 7 |
-
|
| 8 |
-
|
| 9 |
-
|
| 10 |
-
|
| 11 |
-
|
| 12 |
-
|
| 13 |
-
|
| 14 |
-
|
| 15 |
-
|
| 16 |
-
|
| 17 |
-
|
| 18 |
-
|
| 19 |
-
|
| 20 |
-
|
| 21 |
-
|
| 22 |
-
|
| 23 |
-
|
| 24 |
-
|
| 25 |
-
|
| 26 |
-
|
| 27 |
-
input_wav =
|
| 28 |
-
|
| 29 |
-
|
| 30 |
-
|
| 31 |
-
|
| 32 |
-
|
| 33 |
-
|
| 34 |
-
|
| 35 |
-
|
| 36 |
-
|
| 37 |
-
|
| 38 |
-
|
| 39 |
-
|
| 40 |
-
|
| 41 |
-
|
| 42 |
-
|
| 43 |
-
|
| 44 |
-
|
| 45 |
-
|
| 46 |
-
|
| 47 |
-
|
| 48 |
-
|
| 49 |
-
|
| 50 |
-
("
|
| 51 |
-
|
| 52 |
-
|
| 53 |
-
|
| 54 |
-
|
| 55 |
-
|
| 56 |
-
|
| 57 |
-
|
| 58 |
-
|
| 59 |
-
|
| 60 |
-
|
| 61 |
-
|
| 62 |
-
|
| 63 |
-
|
| 64 |
-
|
| 65 |
-
|
| 66 |
-
|
| 67 |
-
|
| 68 |
-
|
| 69 |
-
|
| 70 |
-
|
| 71 |
-
|
| 72 |
-
|
| 73 |
-
|
| 74 |
-
|
| 75 |
-
|
| 76 |
-
|
| 77 |
-
|
| 78 |
-
|
| 79 |
-
|
| 80 |
-
|
| 81 |
-
|
| 82 |
-
|
| 83 |
-
|
| 84 |
-
|
| 85 |
-
|
| 86 |
-
|
| 87 |
-
|
| 88 |
-
|
| 89 |
-
|
| 90 |
-
|
| 91 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import gradio as gr
|
| 2 |
+
import numpy as np
|
| 3 |
+
import torch
|
| 4 |
+
import torchaudio
|
| 5 |
+
import re
|
| 6 |
+
from funasr import AutoModel
|
| 7 |
+
from funasr.utils.postprocess_utils import rich_transcription_postprocess
|
| 8 |
+
|
| 9 |
+
model = AutoModel(
|
| 10 |
+
model="FunAudioLLM/SenseVoiceSmall",
|
| 11 |
+
vad_model="iic/speech_fsmn_vad_zh-cn-16k-common-pytorch",
|
| 12 |
+
vad_kwargs={"max_single_segment_time": 30000},
|
| 13 |
+
hub="hf",
|
| 14 |
+
trust_remote_code=True,
|
| 15 |
+
device="cuda:0" if torch.cuda.is_available() else "cpu",
|
| 16 |
+
)
|
| 17 |
+
|
| 18 |
+
|
| 19 |
+
def model_inference(
|
| 20 |
+
audio_input, language, output_format, use_itn, merge_vad, merge_length, ban_emo_unk
|
| 21 |
+
):
|
| 22 |
+
if audio_input is None:
|
| 23 |
+
return "错误:请上传或录制音频。"
|
| 24 |
+
fs, input_wav = audio_input
|
| 25 |
+
input_wav = input_wav.astype(np.float32) / np.iinfo(np.int16).max
|
| 26 |
+
if len(input_wav.shape) > 1:
|
| 27 |
+
input_wav = input_wav.mean(-1)
|
| 28 |
+
if fs != 16000:
|
| 29 |
+
resampler = torchaudio.transforms.Resample(orig_freq=fs, new_freq=16000)
|
| 30 |
+
input_wav = resampler(torch.from_numpy(input_wav).to(torch.float32)).numpy()
|
| 31 |
+
res = model.generate(
|
| 32 |
+
input=input_wav,
|
| 33 |
+
cache={},
|
| 34 |
+
language=language,
|
| 35 |
+
use_itn=use_itn,
|
| 36 |
+
batch_size_s=60,
|
| 37 |
+
merge_vad=merge_vad,
|
| 38 |
+
merge_length_s=merge_length,
|
| 39 |
+
ban_emo_unk=ban_emo_unk,
|
| 40 |
+
)
|
| 41 |
+
raw_text = res[0].get("text", "未能识别出文本。")
|
| 42 |
+
if output_format == "纯净文本":
|
| 43 |
+
return rich_transcription_postprocess(raw_text)
|
| 44 |
+
return raw_text
|
| 45 |
+
|
| 46 |
+
|
| 47 |
+
html_intro = """<div style="text-align: center; font-family: var(--font-sans);"><h1 style="font-size: 28px;">SenseVoice-Small 语音识别模型</h1><p>SenseVoice 是具有音频理解能力的音频基础模型,包括语音识别(ASR)、语种识别(LID)、语音情感识别(SER)和声学事件分类(AEC)或声学事件检测(AED)。</p><p style="font-size: small; color: #888;">支持 MP3, WAV, FLAC, M4A 等常见音频格式。</p></div>"""
|
| 48 |
+
custom_css = "body, .gradio-container {font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, Oxygen, Ubuntu, Cantarell, 'Open Sans', 'Helvetica Neue', sans-serif;}"
|
| 49 |
+
lang_options = [
|
| 50 |
+
("自动检测", "auto"),
|
| 51 |
+
("中文", "zh"),
|
| 52 |
+
("English", "en"),
|
| 53 |
+
("粤语", "yue"),
|
| 54 |
+
("日本語", "ja"),
|
| 55 |
+
("한국어", "ko"),
|
| 56 |
+
("无语音", "nospeech"),
|
| 57 |
+
]
|
| 58 |
+
|
| 59 |
+
with gr.Blocks(theme=gr.themes.Soft(), css=custom_css) as demo:
|
| 60 |
+
gr.HTML(html_intro)
|
| 61 |
+
with gr.Row():
|
| 62 |
+
with gr.Column(scale=2):
|
| 63 |
+
audio_inputs = gr.Audio(label="上传/录制音频")
|
| 64 |
+
language_inputs = gr.Dropdown(
|
| 65 |
+
choices=lang_options, value="auto", label="源语言"
|
| 66 |
+
)
|
| 67 |
+
output_format_dropdown = gr.Dropdown(
|
| 68 |
+
choices=["纯净文本", "原始富文本", "Emoji 格式"],
|
| 69 |
+
value="纯净文本",
|
| 70 |
+
label="输出格式",
|
| 71 |
+
)
|
| 72 |
+
with gr.Accordion("高级设置", open=False):
|
| 73 |
+
use_itn_checkbox = gr.Checkbox(value=True, label="自动添加标点与格式化")
|
| 74 |
+
merge_vad_checkbox = gr.Checkbox(value=True, label="优化长音频断句")
|
| 75 |
+
merge_length_slider = gr.Slider(
|
| 76 |
+
minimum=5, maximum=30, value=15, step=1, label="断句最大长度(秒)"
|
| 77 |
+
)
|
| 78 |
+
ban_emo_unk_checkbox = gr.Checkbox(value=False, label="强制情感分类")
|
| 79 |
+
fn_button = gr.Button("开始识别", variant="primary")
|
| 80 |
+
with gr.Column(scale=3):
|
| 81 |
+
text_outputs = gr.Textbox(label="识别结果", lines=25, show_copy_button=True)
|
| 82 |
+
fn_button.click(
|
| 83 |
+
fn=model_inference,
|
| 84 |
+
inputs=[
|
| 85 |
+
audio_inputs,
|
| 86 |
+
language_inputs,
|
| 87 |
+
output_format_dropdown,
|
| 88 |
+
use_itn_checkbox,
|
| 89 |
+
merge_vad_checkbox,
|
| 90 |
+
merge_length_slider,
|
| 91 |
+
ban_emo_unk_checkbox,
|
| 92 |
+
],
|
| 93 |
+
outputs=text_outputs,
|
| 94 |
+
api_name="model_inference",
|
| 95 |
+
)
|
| 96 |
+
|
| 97 |
+
demo.launch(server_name="0.0.0.0", server_port=7860)
|
requirements.txt
CHANGED
|
@@ -1,5 +1,8 @@
|
|
| 1 |
-
|
| 2 |
-
modelscope
|
| 3 |
-
torch
|
| 4 |
torchaudio
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 5 |
gradio
|
|
|
|
| 1 |
+
torch>=1.13
|
|
|
|
|
|
|
| 2 |
torchaudio
|
| 3 |
+
funasr>=1.2.0
|
| 4 |
+
huggingface_hub
|
| 5 |
+
modelscope
|
| 6 |
+
numpy<2.0
|
| 7 |
+
librosa
|
| 8 |
gradio
|