Hugging Face's logo

Spaces:
MohamedRashad
/
Arabic-Whisper-CodeSwitching-Edition
Running on Zero

App Files Community
Arabic-Whisper-CodeSwitching-Edition
File size: 4,758 Bytes 
b2b8bd5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
 
import torch
import spaces
import gradio as gr
from pytube import YouTube
from transformers import pipeline, WhisperProcessor, WhisperForConditionalGeneration
from transformers.pipelines.audio_utils import ffmpeg_read

import tempfile
import os

MODEL_NAME = "MohamedRashad/Arabic-Whisper-CodeSwitching-Edition"
BATCH_SIZE = 8
FILE_LIMIT_MB = 1000*3
YT_LENGTH_LIMIT_S = 60*60*3  # limit to 3 hour YouTube files

device = 0 if torch.cuda.is_available() else "cpu"

processor = WhisperProcessor.from_pretrained(MODEL_NAME)
model = WhisperForConditionalGeneration.from_pretrained(MODEL_NAME, torch_dtype=torch.bfloat16)

pipe = pipeline(
    task="automatic-speech-recognition",
    model=model,
    tokenizer=processor.tokenizer,
    feature_extractor=processor.feature_extractor,
    chunk_length_s=30,
    device=device,
)

@spaces.GPU(120)
def transcribe(inputs):
    if inputs is None:
        raise gr.Error("No audio file submitted! Please upload or record an audio file before submitting your request.")

    text = pipe(inputs, batch_size=BATCH_SIZE, generate_kwargs={"task": "transcribe", "language": "arabic"}, return_timestamps=True)["text"]
    return  text


def _return_yt_html_embed(yt_url):
    video_id = YouTube(yt_url).video_id
    HTML_str = (
        f'<center> <iframe width="500" height="320" src="https://www.youtube.com/embed/{video_id}"> </iframe>'
        " </center>"
    )
    return HTML_str

def download_yt_audio(yt_url, filename):
    yt = YouTube(yt_url)
    
    if yt.length > YT_LENGTH_LIMIT_S:
        raise gr.Error("YouTube video is too long! Please upload a video that is less than 1 hour long.")
    
    stream = yt.streams.filter(only_audio=True).first()
    stream.download(filename=filename)

def seconds_to_timestamp(seconds):
    total_seconds = int(seconds)
    hours = total_seconds // 3600
    minutes = (total_seconds % 3600) // 60
    remaining_seconds = seconds % 60
    return f"{hours:02d}:{minutes:02d}:{remaining_seconds:06.3f}"


def chunks_to_subtitle(chunks):
    subtitle = ""
    for chunk in chunks:
        start = seconds_to_timestamp(chunk["timestamp"][0])
        end = seconds_to_timestamp(chunk["timestamp"][1])
        text = chunk["text"]
        subtitle += f"{start} --> {end}\n{text}\n\n"
    return subtitle

@spaces.GPU(120)
def yt_transcribe(yt_url):
    html_embed_str = _return_yt_html_embed(yt_url)

    with tempfile.TemporaryDirectory() as tmpdirname:
        filepath = os.path.join(tmpdirname, "video.mp4")
        download_yt_audio(yt_url, filepath)
        with open(filepath, "rb") as f:
            inputs = f.read()

    inputs = ffmpeg_read(inputs, pipe.feature_extractor.sampling_rate)
    inputs = {"array": inputs, "sampling_rate": pipe.feature_extractor.sampling_rate}

    output = pipe(inputs, batch_size=BATCH_SIZE, generate_kwargs={"task": "transcribe", "language": "arabic"}, return_timestamps=True)
    subtitle = chunks_to_subtitle(output["chunks"])

    return html_embed_str, subtitle


demo = gr.Blocks()

mf_transcribe = gr.Interface(
    fn=transcribe,
    inputs=[
        gr.Audio(sources="microphone", type="filepath"),
    ],
    outputs="text",
    title="Whisper Large V3: Transcribe Audio",
    description=(
        "Transcribe long-form microphone or audio inputs with the click of a button! Demo uses the"
        f" checkpoint [{MODEL_NAME}](https://huggingface.co/{MODEL_NAME}) and 🤗 Transformers to transcribe audio files"
        " of arbitrary length."
    ),
    allow_flagging="never",
)

file_transcribe = gr.Interface(
    fn=transcribe,
    inputs=[
        gr.Audio(sources="upload", type="filepath", label="Audio file"),
    ],
    outputs="text",
    title="Whisper Large V3: Transcribe Audio",
    description=(
        "Transcribe long-form microphone or audio inputs with the click of a button! Demo uses the"
        f" checkpoint [{MODEL_NAME}](https://huggingface.co/{MODEL_NAME}) and 🤗 Transformers to transcribe audio files"
        " of arbitrary length."
    ),
    allow_flagging="never",
)

yt_transcribe_demo = gr.Interface(
    fn=yt_transcribe,
    inputs=[
        gr.Textbox(lines=1, placeholder="Paste the URL to a YouTube video here", label="YouTube URL"),
    ],
    outputs=["html", "text"],
    title="Whisper Large V3: Transcribe YouTube",
    description=(
        "Transcribe long-form YouTube videos with the click of a button! Demo uses the checkpoint"
        f" [{MODEL_NAME}](https://huggingface.co/{MODEL_NAME}) and 🤗 Transformers to transcribe video files of"
        " arbitrary length."
    ),
    allow_flagging="never",
)

with demo:
    gr.TabbedInterface([mf_transcribe, file_transcribe, yt_transcribe_demo], ["Microphone", "Audio file", "YouTube"])

demo.queue().launch(share=True)