File size: 2,128 Bytes
9fbe6a6
 
 
 
 
17f9ae1
9fbe6a6
 
 
 
 
 
 
 
17f9ae1
 
9fbe6a6
17f9ae1
 
9fbe6a6
 
 
 
 
 
 
 
 
 
 
 
e926ef7
17f9ae1
 
 
 
9fbe6a6
 
 
 
 
e926ef7
 
 
 
9fbe6a6
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
import soundfile as sf
import torch
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
import gradio as gr
import sox
import os

def convert(inputfile, outfile):
    sox_tfm = sox.Transformer()
    sox_tfm.set_output_format(
        file_type="wav", channels=1, encoding="signed-integer", rate=16000, bits=16
    )
    sox_tfm.build(inputfile, outfile)


api_token = os.getenv("API_TOKEN")
model_name = "indonesian-nlp/wav2vec2-indonesian-javanese-sundanese"
processor = Wav2Vec2Processor.from_pretrained(model_name, use_auth_token=api_token)
model = Wav2Vec2ForCTC.from_pretrained(model_name, use_auth_token=api_token)
def parse_transcription(wav_file):
    filename = wav_file.name.split('.')[0]
    convert(wav_file.name, filename + "16k.wav")
    speech, _ = sf.read(filename + "16k.wav")
    input_values = processor(speech, sampling_rate=16_000, return_tensors="pt").input_values
    logits = model(input_values).logits
    predicted_ids = torch.argmax(logits, dim=-1)
    transcription = processor.decode(predicted_ids[0], skip_special_tokens=True)
    return transcription



output = gr.outputs.Textbox(label="The transcript")

input_ = gr.inputs.Audio(source="microphone", type="file")

gr.Interface(parse_transcription, inputs=input_,  outputs=[output],
             analytics_enabled=False,
             show_tips=False,
             theme='huggingface',
             layout='vertical',
             title="Multilingual Speech Recognition for Indonesian Languages",
             description="Speech Recognition Live Demo for Indonesian, Javanese and Sundanese Language",
             article="This demo was built for the project [Multilingual Speech Recognition for Indonesian Languages](https://github.com/indonesian-nlp/multilingual-asr). "
                     "It uses the Wav2Vec2 large model [indonesian-nlp/wav2vec2-indonesian-javanese-sundanese](https://huggingface.co/indonesian-nlp/wav2vec2-indonesian-javanese-sundanese) "
                     "which was fine-tuned on Indonesian Common Voice, Javanese and Sundanese OpenSLR speech datasets.",
             enable_queue=True).launch( inline=False)