server-data / app.py
vericudebuget's picture
Update app.py
cbd93b0 verified
raw
history blame
12.4 kB
import streamlit as st
from huggingface_hub import HfApi
import os
import json
from datetime import datetime
import cv2
import random
from PIL import Image
import string
import subprocess
import glob
import shutil
from groq import Groq
import tempfile
from pydub import AudioSegment
# Initialize the Hugging Face and Groq APIs
hf_api = HfApi(token=os.getenv("HF_API_TOKEN"))
groq_client = Groq(api_key=os.getenv("GROQ_API_KEY"))
def generate_random_string(length=4):
return ''.join(random.choices(string.ascii_lowercase, k=length))
def add_random_to_filename(filename):
name, ext = os.path.splitext(filename)
random_string = generate_random_string()
return f"{name}-{random_string}{ext}"
def extract_thumbnail(video_path, thumbnail_path):
video = cv2.VideoCapture(video_path)
total_frames = int(video.get(cv2.CAP_PROP_FRAME_COUNT))
random_frame = random.randint(0, total_frames - 1)
video.set(cv2.CAP_PROP_POS_FRAMES, random_frame)
success, frame = video.read()
if success:
cv2.imwrite(thumbnail_path, frame)
video.release()
return success
def save_custom_thumbnail(thumbnail_file, thumbnail_path):
img = Image.open(thumbnail_file)
img.save(thumbnail_path)
return True
def get_video_length(video_path):
video = cv2.VideoCapture(video_path)
fps = video.get(cv2.CAP_PROP_FPS)
total_frames = int(video.get(cv2.CAP_PROP_FRAME_COUNT))
duration = int(total_frames / fps) if fps > 0 else 0
video.release()
return duration
def generate_metadata(video_name, title, description, uploader, file_location, thumbnail_location, subtitle_location, duration):
return {
"fileName": video_name,
"title": title,
"description": description,
"uploader": uploader,
"uploadTimestamp": datetime.now().isoformat(),
"fileLocation": file_location,
"thumbnailLocation": thumbnail_location,
"subtitleLocation": subtitle_location,
"duration": duration,
"views": 0,
"likes": 0
}
def update_index_file(new_metadata_path):
temp_dir = "temp_repo"
# Remove existing temp directory if it exists
if os.path.exists(temp_dir):
shutil.rmtree(temp_dir)
try:
# Clone the Hugging Face repo
subprocess.run('GIT_LFS_SKIP_SMUDGE=1 git clone https://huggingface.co/spaces/vericudebuget/ok4231 ' + temp_dir,
shell=True,
check=True)
# Find all existing JSON metadata files
metadata_dir = os.path.join(temp_dir, 'metadata')
json_files = glob.glob(os.path.join(metadata_dir, '*-index.json'))
base_url = "https://huggingface.co/spaces/vericudebuget/ok4231/raw/main/metadata/"
paths = []
# Collect existing metadata files with timestamps
for f in json_files:
file_timestamp = datetime.now().isoformat() # Get the current timestamp
file_path = f"{base_url}{os.path.basename(f)}"
paths.append({"url": file_path, "timestamp": file_timestamp})
# Add the new metadata file with the current timestamp
new_metadata_filename = os.path.basename(new_metadata_path)
new_full_path = f"{base_url}{new_metadata_filename}"
file_timestamp = datetime.now().isoformat() # Get timestamp for the new metadata file
# Check if the new file is already in the list, if not, add it
if not any(entry['url'] == new_full_path for entry in paths):
paths.append({"url": new_full_path, "timestamp": file_timestamp})
# Sort the paths by timestamp in descending order (latest to oldest)
paths.sort(key=lambda x: x['timestamp'], reverse=True)
# Convert the paths list to a JSON format
index_content = json.dumps(paths, indent=2)
# Write the sorted index to 'video-index.json'
index_path = os.path.join(temp_dir, 'metadata', 'video-index.json')
os.makedirs(os.path.dirname(index_path), exist_ok=True)
with open(index_path, 'w') as f:
f.write(index_content)
# Upload the updated index file to the Hugging Face space
hf_api.upload_file(
path_or_fileobj=index_path,
path_in_repo="metadata/video-index.json",
repo_id="vericudebuget/ok4231",
repo_type="space",
)
finally:
# Clean up by removing the temp directory
if os.path.exists(temp_dir):
shutil.rmtree(temp_dir)
def create_subtitles(video_path): # Renamed from generate_subtitles
with tempfile.NamedTemporaryFile(suffix='.mp3', delete=False) as temp_audio:
# Convert video to mono 128kbps MP3
audio = AudioSegment.from_file(video_path)
audio = audio.set_channels(1).set_frame_rate(44100).set_sample_width(2)
audio.export(temp_audio.name, format='mp3', bitrate='128k')
# Generate subtitles using Groq
with open(temp_audio.name, 'rb') as audio_file:
translation = groq_client.audio.translations.create(
file=(temp_audio.name, audio_file.read()),
model="whisper-large-v3",
response_format="verbose_json",
temperature=0.0
)
# Generate VTT content
vtt_content = "WEBVTT\n\n"
for segment in translation.segments:
start_time = segment['start']
end_time = segment['end']
text = segment['text'].strip()
start_time_vtt = f"{int(start_time // 3600):02}:{int((start_time % 3600) // 60):02}:{start_time % 60:06.3f}"
end_time_vtt = f"{int(end_time // 3600):02}:{int((end_time % 3600) // 60):02}:{end_time % 60:06.3f}"
vtt_content += f"{start_time_vtt} --> {end_time_vtt}\n{text}\n\n"
os.unlink(temp_audio.name) # Clean up temp file
return vtt_content
def upload_video_to_hf(video_file, original_video_name, title, description, uploader, should_generate_subs=False, custom_thumbnail=None):
temp_dir = "temp"
if not os.path.exists(temp_dir):
os.makedirs(temp_dir)
try:
video_name = add_random_to_filename(original_video_name)
video_path = os.path.join(temp_dir, video_name)
base_name = os.path.splitext(video_name)[0]
thumbnail_name = f"{base_name}_thumb.jpg"
thumbnail_path = os.path.join(temp_dir, thumbnail_name)
json_name = f"{base_name}-index.json"
json_path = os.path.join(temp_dir, json_name)
with open(video_path, "wb") as f:
f.write(video_file.read())
if custom_thumbnail:
thumbnail_extracted = save_custom_thumbnail(custom_thumbnail, thumbnail_path)
else:
thumbnail_extracted = extract_thumbnail(video_path, thumbnail_path)
if not thumbnail_extracted:
st.error("Failed to process thumbnail")
return None
video_length = get_video_length(video_path)
# Analyze audio level
audio = AudioSegment.from_file(video_path)
audio_dBFS = audio.dBFS
# Generate and upload subtitles if requested and video is not too long
subtitle_location = ""
if should_generate_subs and video_length <= 3600: # 1 hour in seconds
if audio_dBFS < -90:
subtitle_location = "" # Set to empty if audio is too quiet
else:
try:
vtt_content = create_subtitles(video_path) # Using renamed function
subtitle_name = f"{base_name}.vtt"
subtitle_path = os.path.join(temp_dir, subtitle_name)
with open(subtitle_path, 'w') as f:
f.write(vtt_content)
subtitle_location = f"subtitles/{subtitle_name}"
hf_api.upload_file(
path_or_fileobj=subtitle_path,
path_in_repo=subtitle_location,
repo_id="vericudebuget/ok4231",
repo_type="space",
)
except Exception as e:
st.warning(f"Failed to generate subtitles: {str(e)}")
# Upload video and thumbnail
video_location = f"videos/{video_name}"
hf_api.upload_file(
path_or_fileobj=video_path,
path_in_repo=video_location,
repo_id="vericudebuget/ok4231",
repo_type="space",
)
thumbnail_location = f"thumbnails/{thumbnail_name}"
hf_api.upload_file(
path_or_fileobj=thumbnail_path,
path_in_repo=thumbnail_location,
repo_id="vericudebuget/ok4231",
repo_type="space",
)
# Generate and upload metadata
metadata = generate_metadata(video_name, title, description, uploader, video_location, thumbnail_location, subtitle_location, video_length)
with open(json_path, "w") as f:
json.dump(metadata, f, indent=2)
metadata_location = f"metadata/{json_name}"
hf_api.upload_file(
path_or_fileobj=json_path,
path_in_repo=metadata_location,
repo_id="vericudebuget/ok4231",
repo_type="space",
)
update_index_file(metadata_location)
return metadata
finally:
if os.path.exists(temp_dir):
shutil.rmtree(temp_dir)
# Streamlit app interface
st.title("Upload your video")
st.markdown("---")
uploaded_video = st.file_uploader("Choose video file", type=["mp4", "avi", "mov", "webm", "mkv"])
if uploaded_video:
with st.form("video_details"):
st.write("Video Details")
title = st.text_input("Title", placeholder="Enter video title")
description = st.text_area("Description", placeholder="Enter video description")
uploader = st.text_input("Uploader Name", placeholder="Enter your name")
# Create a temporary file to get video duration
with tempfile.NamedTemporaryFile(delete=False, suffix='.mp4') as temp_video:
temp_video.write(uploaded_video.getvalue())
video_duration = get_video_length(temp_video.name)
os.unlink(temp_video.name) # Clean up temp file
# Subtitle generation toggle, disabled if video is longer than 2 hours
should_generate_subs = st.toggle("Generate Subtitles. - If enabled, the subtitles will automatically be translated into English.", disabled=video_duration > 7200, value=True) # Renamed variable
if video_duration > 1180:
st.warning("Hey there! Just wanted to warn you that uploading pirated movies is not allowed.")
if video_duration > 3600 and should_generate_subs:
st.warning("Warning, for videos longer than an hour, generating subtitles will take some time! Please wait :)")
if video_duration > 7000:
st.warning("Now that's a long video. It will take a long time to upload. Make sure you have the right uploader details!")
custom_thumbnail = st.file_uploader("Upload custom thumbnail (optional)", type=["jpg", "jpeg", "png"])
submit_button = st.form_submit_button("Upload Video")
if submit_button:
if not title or not uploader:
st.error("Please fill in the title and uploader name.")
else:
with st.spinner("Uploading video, generating thumbnail and metadata... This may take some time. Please wait."):
metadata = upload_video_to_hf(
uploaded_video,
uploaded_video.name,
title,
description,
uploader,
should_generate_subs, # Using renamed variable
custom_thumbnail
)
if metadata:
st.success("Upload completed successfully!")
st.json(metadata)
else:
st.info("Please upload a video file to begin.")