|
import gradio as gr |
|
import whisper |
|
from pytube import YouTube |
|
|
|
loaded_model = whisper.load_model("ksoky/whisper-small-km") |
|
current_size = 'small' |
|
def inference(link): |
|
yt = YouTube(link) |
|
path = yt.streams.filter(only_audio=True)[0].download(filename="audio.mp4") |
|
options = whisper.DecodingOptions(without_timestamps=True) |
|
results = loaded_model.transcribe(path) |
|
return results['text'] |
|
|
|
def change_model(size): |
|
if size == current_size: |
|
return |
|
loaded_model = whisper.load_model(size) |
|
current_size = size |
|
|
|
def populate_metadata(link): |
|
yt = YouTube(link) |
|
return yt.thumbnail_url, yt.title |
|
|
|
title="Youtube Whisperer" |
|
description="Speech to text transcription of Youtube videos using fine-tuned OpenAI's Whisper on Khmer" |
|
block = gr.Blocks() |
|
|
|
with block: |
|
gr.HTML( |
|
""" |
|
<div style="text-align: center; max-width: 500px; margin: 0 auto;"> |
|
<div> |
|
<h1>Youtube Whisperer</h1> |
|
</div> |
|
<p style="margin-bottom: 10px; font-size: 94%"> |
|
Speech to text transcription of Youtube videos using OpenAI's Whisper |
|
</p> |
|
</div> |
|
""" |
|
) |
|
with gr.Group(): |
|
with gr.Box(): |
|
|
|
sz = gr.Dropdown(label="Model Size", choices=['small'], value='small') |
|
link = gr.Textbox(label="YouTube Link") |
|
|
|
with gr.Row().style(mobile_collapse=False, equal_height=True): |
|
title = gr.Label(label="Video Title", placeholder="Title") |
|
img = gr.Image(label="Thumbnail") |
|
text = gr.Textbox( |
|
label="Transcription", |
|
placeholder="Transcription Output", |
|
lines=5) |
|
with gr.Row().style(mobile_collapse=False, equal_height=True): |
|
btn = gr.Button("Transcribe") |
|
|
|
|
|
btn.click(inference, inputs=[link], outputs=[text]) |
|
link.change(populate_metadata, inputs=[link], outputs=[img, title]) |
|
sz.change(change_model, inputs=[sz], outputs=[]) |
|
|
|
block.launch(debug=True) |
