import gradio as gr
import os
import subprocess
# install moviepy dependency
moviepy = subprocess.run(["pip", "install", "moviepy"])
ffmpeg = subprocess.run(["pip", "install", "ffmpeg-python"])
pipUpdate = subprocess.run(["pip", "install", "--upgrade", "pip"])
from azure.storage.blob import BlobServiceClient
import AzureBlobStorageVideo
import AzureBlobStorageAudio
from apiTest import videoAnalysis
from Moviepy import extract_audio_from_video
from Moviepy import load_json_output
from Moviepy import get_explosion_segments
from Moviepy import create_final_audio
from Moviepy import save_audio
from Moviepy import without_audio
from Moviepy import combine_video_audio
from Moviepy import save_video
from moviepy.editor import *
import json
def predict_video(input_video, input_audio=None, input_choice="Explosions"):
global filename, file_size # Use the global keyword to refer to the global variables
# Check if the video is available
if input_video is None:
return [None, "Please upload a video"]
filename = input_video.name # Get the uploaded filename
file_size = os.path.getsize(input_video.name) # Get the file size in bytes
# Loop until a valid video is uploaded
if not filename.lower().endswith('.mp4'):
return [None, "Error: Please upload an MP4 video file."]
if file_size > 20 * 1024 * 1024:
return [None, "Error: The upload exceeds file size 16MB. Please upload a smaller file."]
#Initialize blob storage credentials
storage_account_name = "useruploadhuggingface"
storage_account_key = "zhrGpPBX6PVD+krncC4nVF4yoweEku/z2ErVxjLiuu/CjAVKqM5O4xlGWEyuWGxptL3mA1pv/6P4+AStjSjLEQ=="
connection_string = f"DefaultEndpointsProtocol=https;AccountName={storage_account_name};AccountKey={storage_account_key};EndpointSuffix=core.windows.net"
video_container_name = "useruploadhuggingfacevideo"
audio_container_name = "useruploadhuggingfaceaudio"
# 1. Upload user video file to azure blob storage
videoBlobURL = AzureBlobStorageVideo.uploadUserVideoToBlobStorage(input_video, filename)
videoSASToken = AzureBlobStorageVideo.generateSASToken(storage_account_name,video_container_name, filename, storage_account_key)
videoSASURL = AzureBlobStorageVideo.generateSASURL(storage_account_name, video_container_name, filename, videoSASToken)
# 1.1. Upload user audio if available
userAudioInputFlag = False
if input_audio is not None:
userAudioInputFlag = True
else:
if (input_choice == "Explosions"):
input_audio = os.path.join(os.path.dirname(__file__), "audio/1_seconds_haptic_audio.mp3")
print("explosion selected")
elif (input_choice == "Lightning and Thunder"):
input_audio = os.path.join(os.path.dirname(__file__), "audio/8_seconds_Thunder.mp3")
print("lightning and thunder selected")
elif (input_choice == "Vehicle Racing"):
input_audio = os.path.join(os.path.dirname(__file__), "audio/5_seconds_vehicle_audio.mp3")
print("vehicle racing selected")
else:
input_audio = os.path.join(os.path.dirname(__file__), "audio/5_seconds_haptic_videos.mp3")
print("default selected")
"""
Processes the uploaded video (replace with your video analysis logic).
Args:
input_video: The uploaded video file object.
input_audio (optional): The uploaded audio file object (MP3).
Returns:
A list containing the processed video and a message string.
"""
responseQueryText = videoAnalysis(videoSASURL, videoSASToken, input_choice)
# IF method returns error: run analysis again
if responseQueryText == """{"error":{"code":"InvalidRequest","message":"Value for indexName is invalid."}}""":
responseQueryText = videoAnalysis(videoSASURL, videoSASToken, input_choice)
AzureBlobStorageVideo.delete_container('useruploadhuggingfacevideo')
json_data = load_json_output(responseQueryText)
# Extract audio from the video
audio_path = extract_audio_from_video(input_video)
# Get explosion segments
explosion_segments = get_explosion_segments(json_data)
print(input_audio)
# Create final audio
#final_audio = create_final_audio(audio_path, explosion_segments)
final_audio = create_final_audio(audio_path, input_audio, explosion_segments)
# Save enhanced audio
finalAudioPath = "audio/finalAudio.mp3"
save_audio(final_audio, finalAudioPath)
if (userAudioInputFlag == True):
AzureBlobStorageVideo.delete_container('useruploadhuggingfaceaudio')
# Extract video without audio
current_video = without_audio(VideoFileClip(input_video))
# Combine video with final audio
final_video = combine_video_audio(current_video, final_audio)
# Save final video
save_video(final_video, "video/final_enhanced_video.mp4")
finalVideoPath = "video/final_enhanced_video.mp4"
return [finalVideoPath, f"Video enhancement successful"]
css = """
#col-container {
margin: 0 auto;
max-width: 800px;
}
"""
video_1 = os.path.join(os.path.dirname(__file__), "video/test_video.mp4")
audio_1 = os.path.join(os.path.dirname(__file__), "audio/audioTrack.mp3")
search_1 = "Explosions"
with gr.Blocks(css=css) as demo:
with gr.Column(elem_id="col-container"):
gr.HTML("""
Phone brr
Welcome to the Hugging Face Space of Phone brr! We aim to create more immersive content for mobile phones with the use of haptic audio, this demo focuses on working for a very commonly used special effect of explosions hope you enjoy it.
Instructions:
Step 1: Upload your MP4 video.
Step 2: (Optional) Upload an MP3 audio track.
Step 3:(Optional) Choose the instance you want haptics to be added to
Step 4: Click on submit, and We'll analyse the video and suggest explosion timeframes using Azure Cognitive Services.
Step 5: The Haptic Audio will be mixed into the video and enhanced through AI mastering.
Step 6: View and download the final video with haptics.
""")
with gr.Row():
with gr.Column():
video_in = gr.File(label="Upload a Video", file_types=[".mp4"])
with gr.Row():
audio_in = gr.File(label="Optional: Upload an Audio Track", file_types=[".mp3"])
with gr.Column():
choice_in = gr.Dropdown(
["Explosions", "Lightning and Thunder", "Vehicle Racing"],value=callable(""),
label="Choose", info="Haptic Audio will be added for the selected instance in a video",
allow_custom_value=True
)
with gr.Row():
btn_in = gr.Button("Submit", scale=0)
with gr.Column():
video_out = gr.Video(label="Output Video")
with gr.Row():
text_out = gr.Textbox(label="Output Text")
gr.Examples(
examples=[[video_1,audio_1]],
fn=predict_video,
inputs=[video_in, audio_in,choice_in],
outputs=[video_out, text_out],
#cache_examples=True # Cache examples for faster loading
)
with gr.Column():
gr.HTML("""
Audio Library
Explosion Audio Track 1
Explosion Audio Track 2
Explosion Audio Track 3
Explosion Audio Track 4
Explosion Audio Track 5
Vehicle Audio Track 1
Vehicle Audio Track 2
""")
btn_in.click(
fn=predict_video,
inputs=[video_in,audio_in,choice_in],
outputs=[video_out, text_out],
queue=False
)
demo.launch(debug=True)