MatteoFasulo's picture
New [DEV] version
b9adb22
import sys
import csv
from pathlib import Path
import asyncio
import platform
from argparse import Namespace
import edge_tts
import streamlit as st
import pandas as pd
from src.video_creator import ClipMaker
from utils import rgb_to_bgr
result = None
async def generate_video(
model,
tts_voice,
sub_position,
font,
font_color,
font_size,
url,
non_english,
upload_tiktok,
verbose,
video_num,
background_tab,
max_words,
*args,
**kwargs):
args = Namespace(
model=model,
tts=tts_voice.split('|')[0].strip(),
font=font,
font_color=rgb_to_bgr(font_color.lower()),
font_size=font_size,
sub_position=sub_position,
url=url,
non_english=non_english,
upload_tiktok=upload_tiktok,
verbose=verbose,
mp4_background=background_tab,
max_words=max_words
)
async def get_clip(clip, args):
with st.status("Generating video...", expanded=False) as status:
video_creator = ClipMaker(clip=clip, args=args)
status.update(label="Downloading video...")
video_creator.download_background_video()
status.update(label="Loading model...")
video_creator.load_model()
status.update(label="Creating text...")
video_creator.merge_clip_text()
status.update(label="Generating audio...")
await video_creator.text_to_speech()
status.update(label="Generating transcription...")
video_creator.generate_transcription()
status.update(label="Selecting background...")
video_creator.select_background()
status.update(label="Integrating subtitles...")
video_creator.integrate_subtitles()
print('HERE x3')
if upload_tiktok:
status.update(label="Uploading to TikTok...")
video_creator.upload_to_tiktok()
status.update(label="Video generated!",
state="complete", expanded=False)
return str(video_creator.mp4_final_video)
task = [get_clip(clip, args) for clip in video_num]
result = await asyncio.gather(*task)
if len(result) == 1:
return result[0]
else:
return result[-1] # Return the last video generated if multiple videos are generated
@st.cache_data
def csv_to_df(csv_file):
return pd.read_csv(csv_file, sep='|', encoding='utf-8')
@st.cache_data
def df_to_csv(df):
# Save the edited dataframe to the CSV file
df.to_csv("clips.csv", index=False, sep='|')
return df
# Streamlit Config
st.set_page_config(
page_title="Whisper-TikTok",
page_icon="๐Ÿ’ฌ",
layout="wide",
initial_sidebar_state="expanded",
menu_items={
'Get Help': 'https://github.com/MatteoFasulo/Whisper-TikTok',
'Report a bug': "https://github.com/MatteoFasulo/Whisper-TikTok/issues",
'About':
"""
# Whisper-TikTok
Whisper-TikTok is an innovative AI-powered tool that leverages the prowess of Edge TTS, OpenAI-Whisper, and FFMPEG to craft captivating TikTok videos also with a web application interface!
Mantainer: https://github.com/MatteoFasulo
If you find a bug or if you just have questions about the project feel free to reach me at https://github.com/MatteoFasulo/Whisper-TikTok
Any contribution to this project is welcome to improve the quality of work!
"""
}
)
async def main():
st.title("๐Ÿ† Whisper-TikTok ๐Ÿš€")
st.write("Create a TikTok video with text-to-speech of Microsoft Edge's TTS and subtitles of Whisper model.")
st.subheader("Clip Editor", help="Here you can edit the CSV file with the clips data. Copy-and-paste is supported and compatible with Google Sheets, Excel, and others. You can do bulk-editing by dragging the handle on a cell (similar to Excel)!")
st.write("โ„น๏ธ The CSV file is saved automatically when you click the button below. Every time you edit the CSV file, you must click the button to save the changes otherwise they will be lost.")
edited_df = st.data_editor(csv_to_df("clips.csv"),
num_rows="dynamic")
st.button("Save CSV", on_click=df_to_csv, args=(
edited_df,), help="Save the CSV file with the clips")
st.divider()
with st.sidebar:
model = st.selectbox(
"Whisper Model", ["tiny", "base", "small", "medium", "large"], index=2, help="The model used to generate the subtitles. The bigger the model, the better the results, but the slower the generation. The tiny model is recommended for testing purposes. Medium model is enough for good results in many languages.")
with st.expander("โ„น๏ธ How to use"):
st.write(
"""
1. Choose the clip to generate using the dropdown menu.
2. Choose the model to use for the subtitles.
3. Choose the voice to use for the text-to-speech.
4. Choose the background video to use for the clip.
5. Choose the position of the subtitles.
6. Choose the font, font color, and font size for the subtitles.
7. Check the "Non-english" checkbox if you want to generate a clip in a non-english language.
8. Check the "Upload to TikTok" checkbox if you want to upload the clip to TikTok using the TikTok session cookie. For this step it is required to have a TikTok account and to be logged in on your browser. Then the required cookies.txt file can be generated using the guide specified in the README. The cookies.txt file must be placed in the root folder of the project.
""")
LEFT, RIGHT = st.columns(2)
with LEFT:
st.subheader("General settings")
tts_voice = st.selectbox(
"TTS Voice",
[f"{i['ShortName']} | {i['Gender']} | Tags: {i['VoiceTag']['VoicePersonalities']}" for i in await edge_tts.list_voices()], index=111, help="The voice used to generate the audio. The voice must be in the same language as the subtitles."
)
left, mid, right = st.columns(3)
with left:
# Subtitle font
font = st.selectbox(
"Subtitle font", ["Lexend Bold", "Lexend Regular", "Arial", "Roboto", "Big Condensed Black"], index=0, help="The font used for the subtitles.")
with mid:
# Subtitle font size
font_size = st.slider(
"Subtitle font size", 15, 50, 21, help="The font size for the subtitles. It is recommended to use a font size between 18 and 21.")
with right:
# Subtitle font color
font_color = st.color_picker(
"Subtitle font color", "#fff000", help="The color of the subtitles.")
# Subtitle position
left, right = st.columns(2)
with left:
sub_position = st.slider(
"Subtitle alignment (position)", 1, 9, 5, help="The position of the subtitles. 1 is the bottom left corner, 5 is the center, 9 is the top right corner. This is the alignment feature of FFMPEG subtitles.")
with right:
max_words = st.number_input(
"Maximum number of words per line", min_value=2, max_value=5, value=2, step=1, help="The maximum number of words per line for the subtitles. This is the feature for stable whisper model. It is recommended to use a value between 2 and 3.")
# Background Video URL
url = st.text_input(
"URL Background Video", "https://www.youtube.com/watch?v=dQw4w9WgXcQ", help="The URL of the background video to use for the TikTok video", placeholder="https://www.youtube.com/watch?v=intRX7BRA90")
left, mid, right = st.columns(3)
with left:
# Non-english
non_english = st.checkbox(
"Non-english", help="Check this if you want to generate a video in a non-english language")
with mid:
# Upload to TikTok
upload_tiktok = st.checkbox(
"Upload to TikTok", help="Upload the video to TikTok using the TikTok session cookie. For this step it is required to have a TikTok account and to be logged in on your browser. Then the required cookies.txt file can be generated using this guide (https://github.com/kairi003/Get-cookies.txt-LOCALLY). The cookies.txt file must be placed in the root folder of the project.")
with right:
# Verbose
verbose = st.checkbox(
"Verbose", help="Print the output of the commands used to create the video on your terminal. Useful for debugging.")
st.divider()
st.subheader("Video settings")
st.write("CSV file with the clips")
# Get the list of files in "background"
folder_path = Path("background").absolute()
files = folder_path.glob('*.mp4')
files = [file.name for file in files]
# Create a Dropdown with the list of files
background_tab = st.selectbox(
"Your Backgrounds", files, index=0, help="The background video to use for the clip")
# Choose which clip to generate the video for
with open('clips.csv', 'r', encoding='utf-8') as csvfile:
clips = csv.DictReader(csvfile, delimiter='|')
video_num = st.multiselect(
"Video",
options=clips,
format_func=lambda video: f"{video['series']} - {video['part']}",
help="The clip to generate. If you want to generate multiple clips, select them as a multiselect."
)
if st.button("Generate Clip"):
if not video_num:
st.error("You must select at least one clip to generate")
return
global result
result = await generate_video(model, tts_voice, sub_position, font, font_color, font_size,
url, non_english, upload_tiktok, verbose, video_num, background_tab, max_words)
with RIGHT:
if result:
# Put the video in a container
st.video(result)
if __name__ == "__main__":
if platform.system() == 'Windows':
asyncio.set_event_loop_policy(asyncio.WindowsSelectorEventLoopPolicy())
loop = asyncio.new_event_loop()
loop.run_until_complete(main())
loop.close()
sys.exit(0)