File size: 2,663 Bytes
3f44072 de18ebe 9e20b12 d78cd77 9e20b12 8983ff3 3f44072 d5bc6cb 8983ff3 d78cd77 9e20b12 d5bc6cb 8983ff3 d5bc6cb 3f44072 b66750f d78cd77 9e20b12 d78cd77 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 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 |
import gradio as gr
from pyannote.audio import Pipeline
from transformers import pipeline
asr = pipeline(
"automatic-speech-recognition",
model="facebook/wav2vec2-large-960h-lv60-self",
feature_extractor="facebook/wav2vec2-large-960h-lv60-self",
)
pipeline1 = Pipeline.from_pretrained("pyannote/speaker-segmentation")
def diarization(file_input,mic_input,selection):
mic_path = None if mic_input is None else mic_input.name
audio = file_input if selection == "Upload" else mic_path
if audio is None:
return "Please check your inputs!", ""
speaker_output = pipeline1(audio)
text_output = asr(audio,return_timestamps="word")
full_text = text_output['text'].lower()
chunks = text_output['chunks']
diarized_output = ""
i = 0
for turn, _, speaker in speaker_output.itertracks(yield_label=True):
diarized = ""
while i < len(chunks):
time_index = chunks[i]['timestamp'][1]
if time_index >= turn.start and time_index <= turn.end:
diarized += chunks[i]['text'].lower() + ' '
if time_index >= turn.end: break
i += 1
diarized_output += "{} said '{}' from {:.3f} to {:.3f}\n".format(speaker,diarized,turn.start,turn.end)
return diarized_output, full_text
title = "Speech Recognition with Speaker Diarization"
description = "Speaker Diarization is the act of attributing parts of the audio recording to different speakers. This space aims to distinguish the speakers and apply speech-to-text from a given input audio file. Pre-trained models from Pyannote[1] for the Speaker Diarization and [2]."
article = "<p style='text-align: center'><a href='https://github.com/pyannote/pyannote-audio' target='_blank'>[1] Pyannote - Speaker Diarization model</a></p>"
inputs = [gr.inputs.Audio(source="upload", type="filepath", label="Upload your audio file here:", optional=True),
gr.inputs.Audio(source="microphone", type="file",label="Or use your Microphone:", optional=True),
gr.inputs.Radio(["Upload","Microphone"], type="value", label="Select which input:")]
outputs = [gr.outputs.Textbox(type="auto", label="Diarized Output"),
gr.outputs.Textbox(type="auto",label="Full ASR Text for comparison")]
examples = [["test_audio1.wav",None,"Upload"],
["test_audio2.wav",None,"Upload"]]
app = gr.Interface(fn=diarization,
inputs=inputs,
outputs=outputs,
examples=examples,
title=title,
description=description,
article=article,
allow_flagging=False)
app.launch() |