File size: 2,820 Bytes
382e37a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
60
61
62
63
64
65
66
67
68
import gradio as gr
import torchaudio
import torch
from transformers import Wav2Vec2Processor, Wav2Vec2ForCTC
from transformers import Speech2Text2Processor, Speech2Text2ForConditionalGeneration
from transformers import Wav2Vec2Processor, Wav2Vec2ForSequenceClassification

# Load the models
asr_model = Wav2Vec2ForCTC.from_pretrained("facebook/mms-1b-all")
asr_processor = Wav2Vec2Processor.from_pretrained("facebook/mms-1b-all")

tts_model = Speech2Text2ForConditionalGeneration.from_pretrained("facebook/mms-tts")
tts_processor = Speech2Text2Processor.from_pretrained("facebook/mms-tts")

lid_model = Wav2Vec2ForSequenceClassification.from_pretrained("facebook/mms-lid-1024")
lid_processor = Wav2Vec2Processor.from_pretrained("facebook/mms-lid-1024")

# ASR Function
def asr_transcribe(audio):
    inputs = asr_processor(audio, sampling_rate=16000, return_tensors="pt", padding=True)
    with torch.no_grad():
        logits = asr_model(**inputs).logits
    predicted_ids = torch.argmax(logits, dim=-1)
    transcription = asr_processor.batch_decode(predicted_ids)
    return transcription[0]

# TTS Function
def tts_synthesize(text):
    inputs = tts_processor(text, return_tensors="pt", padding=True)
    with torch.no_grad():
        generated_ids = tts_model.generate(**inputs)
    audio = tts_processor.batch_decode(generated_ids, skip_special_tokens=True)
    return audio[0]

# Language ID Function
def identify_language(audio):
    inputs = lid_processor(audio, sampling_rate=16000, return_tensors="pt", padding=True)
    with torch.no_grad():
        logits = lid_model(**inputs).logits
    predicted_ids = torch.argmax(logits, dim=-1)
    language = lid_processor.batch_decode(predicted_ids)
    return language[0]

# Define the Gradio interfaces
with gr.Blocks() as demo:
    with gr.Tab("ASR"):
        gr.Markdown("## Automatic Speech Recognition (ASR)")
        audio_input = gr.Audio(source="microphone", type="numpy")
        text_output = gr.Textbox(label="Transcription")
        gr.Button("Clear", clear_audio_input)
        gr.Button("Submit", fn=asr_transcribe, inputs=audio_input, outputs=text_output)
        
    with gr.Tab("TTS"):
        gr.Markdown("## Text-to-Speech (TTS)")
        text_input = gr.Textbox(label="Text")
        audio_output = gr.Audio(label="Audio Output")
        gr.Button("Clear", clear_text_input)
        gr.Button("Submit", fn=tts_synthesize, inputs=text_input, outputs=audio_output)

    with gr.Tab("Language ID"):
        gr.Markdown("## Language Identification (LangID)")
        audio_input = gr.Audio(source="microphone", type="numpy")
        language_output = gr.Textbox(label="Identified Language")
        gr.Button("Clear", clear_audio_input)
        gr.Button("Submit", fn=identify_language, inputs=audio_input, outputs=language_output)

demo.launch()