2023-03-09 11:46:50 +00:00
|
|
|
import gradio as gr
|
|
|
|
import speech_recognition as sr
|
2023-03-09 19:33:00 +00:00
|
|
|
|
2023-03-09 11:46:50 +00:00
|
|
|
input_hijack = {
|
|
|
|
'state': False,
|
|
|
|
'value': ["", ""]
|
|
|
|
}
|
|
|
|
|
|
|
|
|
2023-03-12 20:03:07 +00:00
|
|
|
def do_stt(audio, text_state=""):
|
2023-03-09 11:46:50 +00:00
|
|
|
transcription = ""
|
|
|
|
r = sr.Recognizer()
|
2023-03-12 20:03:07 +00:00
|
|
|
|
|
|
|
# Convert to AudioData
|
|
|
|
audio_data = sr.AudioData(sample_rate=audio[0], frame_data=audio[1], sample_width=4)
|
2023-03-09 11:46:50 +00:00
|
|
|
|
|
|
|
try:
|
2023-03-12 20:03:07 +00:00
|
|
|
transcription = r.recognize_whisper(audio_data, language="english", model="base.en")
|
2023-03-09 11:46:50 +00:00
|
|
|
except sr.UnknownValueError:
|
|
|
|
print("Whisper could not understand audio")
|
|
|
|
except sr.RequestError as e:
|
2023-03-09 19:33:00 +00:00
|
|
|
print("Could not request results from Whisper", e)
|
2023-03-09 11:46:50 +00:00
|
|
|
|
|
|
|
input_hijack.update({"state": True, "value": [transcription, transcription]})
|
2023-03-12 20:03:07 +00:00
|
|
|
|
|
|
|
text_state += transcription + " "
|
|
|
|
return text_state, text_state
|
2023-03-09 11:46:50 +00:00
|
|
|
|
|
|
|
|
2023-03-09 20:03:49 +00:00
|
|
|
def update_hijack(val):
|
|
|
|
input_hijack.update({"state": True, "value": [val, val]})
|
|
|
|
return val
|
|
|
|
|
|
|
|
|
2023-03-09 11:46:50 +00:00
|
|
|
def ui():
|
2023-03-12 20:03:07 +00:00
|
|
|
tr_state = gr.State(value="")
|
|
|
|
output_transcription = gr.Textbox(label="STT-Input",
|
|
|
|
placeholder="Speech Preview. Click \"Generate\" to send",
|
|
|
|
interactive=True)
|
|
|
|
output_transcription.change(fn=update_hijack, inputs=[output_transcription], outputs=[tr_state])
|
|
|
|
with gr.Row():
|
|
|
|
audio = gr.Audio(source="microphone")
|
|
|
|
transcribe_button = gr.Button(value="Transcribe")
|
|
|
|
transcribe_button.click(do_stt, inputs=[audio, tr_state], outputs=[output_transcription, tr_state])
|