Hugging Face's logo

Spaces:
Bishan
/
Speech_To_Text_Odia
Runtime error

App Files Community
Speech_To_Text_Odia
File size: 2,661 Bytes 
2fa92c9
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
5c7d3a4
 
 
 
 
ee77069
 
 
4cb82b4
 
ee77069
 
 
4cb82b4
 
cb47b32
2fa92c9
8df9495
2fa92c9
 
 
 
 
 
 
 
 
 
 
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
69
70
71
72
73
74
75
76
 
import soundfile as sf
import torch
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor,Wav2Vec2ProcessorWithLM
import gradio as gr
import sox
import subprocess


def read_file_and_process(wav_file):
    filename = wav_file.split('.')[0]
    filename_16k = filename + "16k.wav"
    resampler(wav_file, filename_16k)
    speech, _ = sf.read(filename_16k)
    inputs = processor(speech, sampling_rate=16_000, return_tensors="pt", padding=True)
    
    return inputs


def resampler(input_file_path, output_file_path):
    command = (
        f"ffmpeg -hide_banner -loglevel panic -i {input_file_path} -ar 16000 -ac 1 -bits_per_raw_sample 16 -vn "
        f"{output_file_path}"
    )
    subprocess.call(command, shell=True)


def parse_transcription_with_lm(logits):
    result = processor_with_LM.batch_decode(logits.cpu().numpy())
    text = result.text
    transcription = text[0].replace('<s>','')
    return transcription

def parse_transcription(logits):
    predicted_ids = torch.argmax(logits, dim=-1)
    transcription = processor.decode(predicted_ids[0], skip_special_tokens=True)
    return transcription

def parse(wav_file, applyLM):
    input_values = read_file_and_process(wav_file)
    with torch.no_grad():
        logits = model(**input_values).logits
   
    if applyLM:
        return parse_transcription_with_lm(logits)
    else:
        return parse_transcription(logits)


# processor = Wav2Vec2Processor.from_pretrained("Harveenchadha/vakyansh-wav2vec2-odia-orm-100")
# processor_with_LM = Wav2Vec2ProcessorWithLM.from_pretrained("Harveenchadha/vakyansh-wav2vec2-odia-orm-100")
# model = Wav2Vec2ForCTC.from_pretrained("Harveenchadha/vakyansh-wav2vec2-odia-orm-100")


processor = Wav2Vec2Processor.from_pretrained("Harveenchadha/odia_large_wav2vec2")
processor_with_LM = Wav2Vec2ProcessorWithLM.from_pretrained("Harveenchadha/odia_large_wav2vec2")
model = Wav2Vec2ForCTC.from_pretrained("Harveenchadha/odia_large_wav2vec2")


# processor = Wav2Vec2Processor.from_pretrained("theainerd/wav2vec2-large-xlsr-53-odia")
# processor_with_LM = Wav2Vec2ProcessorWithLM.from_pretrained("theainerd/wav2vec2-large-xlsr-53-odia")
# model = Wav2Vec2ForCTC.from_pretrained("theainerd/wav2vec2-large-xlsr-53-odia")



    
input_ = gr.Audio(source="upload", type="filepath")  
txtbox = gr.Textbox(
            label="Output from model will appear here:",
            lines=5
        )
chkbox = gr.Checkbox(label="Apply LM", value=False)


gr.Interface(parse, inputs = [input_, chkbox],  outputs=txtbox,
             streaming=True, interactive=True,
             analytics_enabled=False, show_tips=False, enable_queue=True).launch(inline=False);