Spaces:
Running
Running
File size: 12,564 Bytes
02cb85f 1dae8a0 02cb85f 1dae8a0 02cb85f 1dae8a0 02cb85f 1dae8a0 02cb85f 1dae8a0 02cb85f 7af11d6 02cb85f 1dae8a0 7af11d6 02cb85f 27131f1 02cb85f |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 |
import json
import os
import shutil
import subprocess
import sys
import time
import math
import cv2
import requests
from pydub import AudioSegment
import numpy as np
from dotenv import load_dotenv
import gradio as gr
from gradio_client import Client, file
# Function to get a friendly name from an audio file name
def get_friendly_name(filename):
return os.path.splitext(filename)[0].capitalize()
# Get audio files and their friendly names
audio_files_dir = "audio_folder" # Path to your audio folder
audio_files = [(get_friendly_name(f), f) for f in os.listdir(audio_files_dir) if f.endswith(".mp3") or f.endswith(".wav")]
# Load environment variables
load_dotenv(override=True)
LEMONFOX_API_KEY = os.getenv("LEMONFOX_API_KEY")
def parse(narration):
data = []
narrations = []
lines = narration.split("\n")
for line in lines:
if line.startswith('Narrator: '):
text = line.replace('Narrator: ', '')
data.append({
"type": "text",
"content": text.strip('"'),
})
narrations.append(text.strip('"'))
elif line.startswith('['):
background = line.strip('[]')
data.append({
"type": "image",
"description": background,
})
return data, narrations
def create(data, output_folder, audio_file):
if not os.path.exists(output_folder):
os.makedirs(output_folder)
# Initialize Gradio Client
client = Client("tonyassi/voice-clone")
audio_files_dir = "audio_folder" # Path to your audio folder
for element in data:
if element["type"] != "text":
continue
# Make prediction using the provided API
audio_file_path = os.path.join(audio_files_dir, audio_file)
result = client.predict(
text=element["content"],
audio=file(audio_file_path) # Include reference style audio for API
)
# Move the response audio file to the output folder
temp_dir = os.path.dirname(result)
response_file_path = os.path.join(output_folder, f"narration_{len(os.listdir(output_folder)) + 1}.wav")
shutil.move(result, response_file_path)
print(f"Audio file generated for '{element['content']}' saved at: {response_file_path}")
def generate(prompt, output_file, size="576x1024"):
url = 'https://api.lemonfox.ai/v1/images/generations'
headers = {
'Authorization': LEMONFOX_API_KEY,
'Content-Type': 'application/json'
}
data = {
'prompt': prompt,
'size': size,
'n': 1
}
try:
response = requests.post(url, json=data, headers=headers)
if response.ok:
response_data = response.json()
if 'data' in response_data and len(response_data['data']) > 0:
image_info = response_data['data'][0]
image_url = image_info['url']
image_response = requests.get(image_url)
with open(output_file, 'wb') as f:
f.write(image_response.content)
else:
print(f"No image data found for prompt: {prompt}")
else:
print(f"Failed to generate image for prompt: {prompt}. Status Code: {response.status_code}")
except Exception as e:
print(f"Error occurred while processing prompt: {prompt}")
print(str(e))
def create_from_data(data, output_dir):
if not os.path.exists(output_dir):
os.makedirs(output_dir)
image_number = 0
for element in data:
if element["type"] != "image":
continue
image_number += 1
image_name = f"image_{image_number}.webp"
generate(element["description"], os.path.join(output_dir, image_name))
def get_audio_duration(audio_file):
return len(AudioSegment.from_file(audio_file))
def resize_image(image, width, height):
aspect_ratio = image.shape[1] / image.shape[0]
if aspect_ratio > (width / height):
new_width = width
new_height = int(width / aspect_ratio)
else:
new_height = height
new_width = int(height * aspect_ratio)
return cv2.resize(image, (new_width, new_height))
def write_text(text, frame, video_writer):
font = cv2.FONT_HERSHEY_SIMPLEX
white_color = (255, 255, 255)
black_color = (0, 0, 0)
thickness = 10
font_scale = 3
border = 5
text_size = cv2.getTextSize(text, font, font_scale, thickness)[0]
text_x = (frame.shape[1] - text_size[0]) // 2
text_y = (frame.shape[0] + text_size[1]) // 2
org = (text_x, text_y)
frame = cv2.putText(frame, text, org, font, font_scale, black_color, thickness + border * 2, cv2.LINE_AA)
frame = cv2.putText(frame, text, org, font, font_scale, white_color, thickness, cv2.LINE_AA)
video_writer.write(frame)
def add_narration_to_video(narrations, input_video, output_dir, output_file, text_color, text_position):
offset = 50
cap = cv2.VideoCapture(input_video)
fourcc = cv2.VideoWriter_fourcc(*'XVID')
temp_video = os.path.join(output_dir, "with_transcript.avi")
out = cv2.VideoWriter(temp_video, fourcc, 60, (int(cap.get(3)), int(cap.get(4))))
full_narration = AudioSegment.empty()
for i, narration in enumerate(narrations):
audio = os.path.join(output_dir, "narrations", f"narration_{i+1}.wav")
duration = get_audio_duration(audio)
narration_frames = math.floor(duration / 2000 * 60)
full_narration += AudioSegment.from_file(audio)
char_count = len(narration.replace(" ", ""))
ms_per_char = duration / char_count
frames_written = 0
words = narration.split(" ")
for w, word in enumerate(words):
word_ms = len(word) * ms_per_char
if i == 0 and w == 0:
word_ms -= offset
if word_ms < 0:
word_ms = 0
for _ in range(math.floor(word_ms/2000*60)):
ret, frame = cap.read()
if not ret:
break
write_text(word, frame, out)
frames_written += 1
for _ in range(narration_frames - frames_written):
ret, frame = cap.read()
out.write(frame)
while out.isOpened():
ret, frame = cap.read()
if not ret:
break
out.write(frame)
temp_narration = os.path.join(output_dir, "narration.wav")
full_narration.export(temp_narration, format="wav")
cap.release()
out.release()
cv2.destroyAllWindows()
ffmpeg_command = [
'ffmpeg',
'-y',
'-i', temp_video,
'-i', temp_narration,
'-map', '0:v',
'-map', '1:a',
'-c:v', 'copy',
'-c:a', 'aac',
'-strict', 'experimental',
os.path.join(output_dir, output_file)
]
subprocess.run(ffmpeg_command, capture_output=True)
os.remove(temp_video)
os.remove(temp_narration)
def create_video(narrations, output_dir, output_file, text_color, text_position): # Add text_color and text_position parameters here
width, height = 1080, 1920
frame_rate = 60
fade_time = 2000
fourcc = cv2.VideoWriter_fourcc(*'XVID')
temp_video = os.path.join(output_dir, "temp_video.avi")
out = cv2.VideoWriter(temp_video, fourcc, frame_rate, (width, height))
image_paths = os.listdir(os.path.join(output_dir, "images"))
image_count = len(image_paths)
for i in range(image_count):
image1 = cv2.imread(os.path.join(output_dir, "images", f"image_{i+1}.webp"))
if i+1 < image_count:
image2 = cv2.imread(os.path.join(output_dir, "images", f"image_{i+2}.webp"))
else:
image2 = cv2.imread(os.path.join(output_dir, "images", f"image_1.webp"))
image1 = resize_image(image1, width, height)
image2 = resize_image(image2, width, height)
narration = os.path.join(output_dir, "narrations", f"narration_{i+1}.wav")
duration = get_audio_duration(narration)
if i > 0:
duration -= fade_time
if i == image_count-1:
duration -= fade_time
for _ in range(math.floor(duration/2000*60)):
vertical_video_frame = np.zeros((height, width, 3), dtype=np.uint8)
vertical_video_frame[:image1.shape[0], :] = image1
out.write(vertical_video_frame)
for alpha in np.linspace(0, 1, math.floor(fade_time/1000*30)):
blended_image = cv2.addWeighted(image1, 1 - alpha, image2, alpha, 0)
vertical_video_frame = np.zeros((height, width, 3), dtype=np.uint8)
vertical_video_frame[:image1.shape[0], :] = blended_image
out.write(vertical_video_frame)
out.release()
cv2.destroyAllWindows()
add_narration_to_video(narrations, temp_video, output_dir, output_file, text_color, text_position) # Pass text_color and text_position here
os.remove(temp_video)
def generate_video(topic, voice_choice):
short_id = str(int(time.time()))
basedir = os.path.join("shorts", short_id)
if not os.path.exists(basedir):
os.makedirs(basedir)
filename = topic.replace("_", " ").replace("/", "_").replace(".", "_")
output_file = f"{filename}.avi"
# Extract the voice file based on voice_choice
voice_file = [file for name, file in audio_files if name == voice_choice][0]
chat_url = 'https://api.lemonfox.ai/v1/chat/completions'
headers = {
'Authorization': f'Bearer {LEMONFOX_API_KEY}',
'Content-Type': 'application/json'
}
payload = {
"model": "mixtral-chat",
"messages": [
{
"role": "system",
"content": "You are a YouTube short video creator."
},
{
"role": "user",
"content": f"""make a short video on: \n\n{topic} Generate 60 seconds to 1 minute of video. You will need to generate a very short description of images for each of the sentences. They will be used for background images. Note that the script will be fed into a text-to-speech engine, so dont use special characters. Respond with a pair of an image description in square brackets and a script below it. Both of them should be on their own lines, as follows: ###
[Description of a background image]
Narrator: "One sentence of narration"
### The short should be 6 sentences maximum."""
}
]
}
response = requests.post(chat_url, json=payload, headers=headers)
if response.status_code == 200:
response_text = response.json()['choices'][0]['message']['content']
response_text = response_text.replace("’", "'").replace("`", "'").replace("…", "...").replace("“", '"').replace("”", '"')
with open(os.path.join(basedir, f"response.txt"), "a") as f:
f.write(response_text + "\n")
data, narrations = parse(response_text)
with open(os.path.join(basedir, f"data.json"), "a") as f:
json.dump(data, f, ensure_ascii=False)
f.write("\n")
print(f"Generating narration for: {topic}...")
create(data, os.path.join(basedir, f"narrations"), voice_file)
print("Generating images...")
create_from_data(data, os.path.join(basedir, f"images"))
print("Generating video...")
create_video(narrations, basedir, output_file, text_color="white", text_position="center") # Pass text_color and text_position here
print("Deleting files and folders...")
os.remove(os.path.join(basedir, "response.txt"))
os.remove(os.path.join(basedir, "data.json"))
shutil.rmtree(os.path.join(basedir, "narrations"))
shutil.rmtree(os.path.join(basedir, "images"))
print(f"DONE! Here's your video: {os.path.join(basedir, output_file)}")
return os.path.join(basedir, output_file)
else:
print(f"Failed to generate script for source material: {topic}. Status Code: {response.status_code}")
return None
iface = gr.Interface(
fn=generate_video,
inputs=["text", gr.Dropdown(choices=[name for name, _ in audio_files], label="Select Voice")],
outputs="video",
css="footer {visibility: hidden}",
description="Generate a free short video. Best for YouTube Shorts, Instagram Reels or TikTok. This is a prototype. If you want better software, please inbox or email me at aheedsajid@gmail.com and do like and [Click here to Donate](https://nowpayments.io/donation/aheed)",
title="Text to Short Video Free"
)
iface.launch() |