main.py

import os
import asyncio
from moviepy.audio.AudioClip import CompositeAudioClip
from moviepy.audio.fx.volumex import volumex
from moviepy.audio.io.AudioFileClip import AudioFileClip
from moviepy.video.io.VideoFileClip import VideoFileClip
from openai import OpenAI
from pydub.silence import detect_silence, split_on_silence, detect_nonsilent
from sydney import SydneyClient
import gtts
from pydub import AudioSegment
from PIL import Image
import base64
from io import BytesIO
import cv2
import os

#here you need to put your OpenAI api key
client = OpenAI(
    api_key = os.getenv('OPENAI_API_KEY'),
)

def query_sydney(prompt):
    loop = asyncio.get_event_loop()
    result = loop.run_until_complete(get_result(prompt))
    return result

async def get_result(title):
    prompt = f"Can you write me a text about {title} in 120 words? Without the sources or anything else, just the text that would be spoken in a presentation about this topic. In English please"
    print("entered get_result")
    sydney = SydneyClient()
    #start conversation
    await sydney.start_conversation()
    print("started conversation")
    #give prompt
    response = await sydney.compose(prompt, format = "blogpost", tone = "informational", length = "short")
    print(response)
    #close conversation
    await sydney.close_conversation()
    print("closed conversation")
    return response

print("Project made by Gherca Darius")
title = input("Enter the title of the video: ")

text = query_sydney(title)
text = "".join([i for i in text if i not in ['`', '´', '’', '‘', '“', '”', '–', '—', '.'] and (i.isdigit() or i.isalpha() or i.isspace())])
# text = text.rstrip('`')
print(text)
# # AUDIO GENERATION
sound1 = gtts.gTTS(text, slow = False)
sound1.save("sound1.mp3")
#
# # AUDIO MANIPULATION
#
print("read file sound1.wav")
audio = AudioSegment.from_file("sound1.mp3", format="mp3")
print("changing playback speed")
audio = audio.speedup(playback_speed=1.25)
print("changed playback speed, exporting")
audio.export("sound2.wav", format="wav")
print("exported")


#IMAGE GENERATION

def generate_image(text):
        response = client.images.generate(prompt = text, model="dall-e-2", n = 5, response_format="b64_json", size="1024x1024", user = "user123")
        json_file = "image_json.json"
        image_data_list = []
        result = []
        for image in response.data:
            image_data_list.append(image.model_dump()["b64_json"])
        image_objects = []
        for i, data in enumerate(image_data_list):
            image_objects.append(Image.open(BytesIO(base64.b64decode(data))))
            image_objects[i].save(f"image{i}.png")
            result.append(f"image{i}.png")
            print("saved image " + f"image{i}.png")
        return result
#
images_names = generate_image(title)
# images_names = ["image0.png", "image1.png", "image2.png", "image3.png", "image4.png"]
#
# #modify image sizes
#
for image in images_names:
    with Image.open(image) as img:
        img = img.resize((1080, 1920))
        img.save(image)
        print(f"resized image {image} successfully")

# # VIDEO GENERATION
#
frame_width, frame_height = 1080, 1920
fps = 30
num_frames = fps * 12
video_title = "video.mp4"
video_writer = cv2.VideoWriter(video_title, cv2.VideoWriter_fourcc(*'mp4v'), fps, (frame_width, frame_height))
print(f"made video file {video_title}")

for image in images_names:
    frame = cv2.imread(image)
    for x in range(num_frames):
        video_writer.write(frame)
    print(f"added image {image} to video for {num_frames} frames")
    # video_writer.write(frame)

video_writer.release()
cv2.destroyAllWindows()
#
# # LINK AUDIO SPEECH TO VIDEO
#
# # clip the sound2.wav to match the length of the video
print("read file sound2.wav")
audio = AudioSegment.from_file("sound2.wav", format="wav")
audio = audio[: 60 * 1000]
audio.export("sound2.wav", format="wav")
print("exported sound2.wav")

print("read file background_audio.wav")
audio = AudioSegment.from_file("background_audio.wav", format="wav")
audio = audio[: 60 * 1000]
audio.export("background_audio.wav", format="wav")
print("exported background_audio.wav")
#
# # create the subtitles for the sound2.wav file based on text
print("opened sound2.wav for subtitles")
audioSub = open("sound2.wav", "rb")
transcript = client.audio.transcriptions.create(
    file=audioSub,
    model="whisper-1",
    response_format="srt"
)
# #write transcript in a srt file
with open("subtitles.srt", "w") as file:
    file.write(transcript)
print("created subtitles")
audioSub.close()
print(transcript)
#
# # load video and audio
#
video_clip = VideoFileClip("video.mp4")
audio_clip = AudioFileClip("sound2.wav")
background_audio = AudioFileClip("background_audio.wav")
volume_factor = 1
# bigger -> louder, smaller -> quieter
audio_clip = volumex(audio_clip, volume_factor)
volume_factor = 0.2
background_audio = volumex(background_audio, volume_factor)
print("composing audio and background music")
final_audio = CompositeAudioClip([audio_clip, background_audio])
print("added audio")
final_clip = video_clip.set_audio(final_audio)
print("set audio")
final_clip.write_videofile("final_video.mp4", codec="libx264", audio_codec="aac")
print("exported successfully without subtitles")
final_clip.close()
#
# # add subtitles to the video
print("started ffmpeg command scripting")
import ffmpeg
(
    ffmpeg
    .input("final_video.mp4")
    .output("final_video_with_subtitles.mp4", vf="subtitles=subtitles.srt:force_style='Fontsize=14, PrimaryColour=&Hffffff&'")
    .run(overwrite_output=True)
)
print("running ffmpeg command")
print("exported successfully with subtitles")
#