File size: 12,138 Bytes
52409f1 d348068 52409f1 b19f57b 52409f1 f405812 d348068 52409f1 b1cd1c8 d68759e 52409f1 d68759e 52409f1 d68759e 52409f1 d68759e 52409f1 b19f57b d68759e c237b6b d68759e c237b6b d68759e 52409f1 b4ce27c 6776e8c 52409f1 2c75576 6776e8c 52409f1 d68759e 6776e8c 2c75576 52409f1 d348068 b1cd1c8 52409f1 b19f57b 52409f1 8e6b896 d68759e 8e6b896 d68759e 8e6b896 b1cd1c8 52409f1 b4ce27c 52409f1 b4ce27c 52409f1 b4ce27c 52409f1 b4ce27c 52409f1 b4ce27c 52409f1 b4ce27c 52409f1 b4ce27c 52409f1 b1cd1c8 52409f1 4cda010 b19f57b b1cd1c8 b19f57b b1cd1c8 52409f1 b19f57b 52409f1 b1cd1c8 52409f1 b1cd1c8 c237b6b b1cd1c8 e75d696 b19f57b b1cd1c8 e75d696 52409f1 b4ce27c b1cd1c8 52409f1 b1cd1c8 52409f1 b1cd1c8 52409f1 b19f57b 52409f1 cee1bed c237b6b 52409f1 291cd48 52409f1 b1cd1c8 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 |
import gradio as gr
from PIL import Image
from moviepy.editor import VideoFileClip, AudioFileClip
import os
from openai import OpenAI
import subprocess
from pathlib import Path
import uuid
import tempfile
import shlex
import shutil
from utils import format_bash_command
HF_API_KEY = os.environ["HF_TOKEN"]
print('caca')
print(os.environ["HF_TOKEN"])
client = OpenAI(
base_url="https://api-inference.huggingface.co/v1/",
api_key=HF_API_KEY
)
allowed_medias = [
".png",
".jpg",
".jpeg",
".tiff",
".bmp",
".gif",
".svg",
".mp3",
".wav",
".ogg",
".mp4",
".avi",
".mov",
".mkv",
".flv",
".wmv",
".webm",
".mpg",
".mpeg",
".m4v",
".3gp",
".3g2",
".3gpp",
]
def get_files_infos(files):
results = []
for file in files:
file_path = Path(file.name)
info = {}
info["size"] = os.path.getsize(file_path)
info["name"] = file_path.name
file_extension = file_path.suffix
if file_extension in (".mp4", ".avi", ".mkv", ".mov"):
info["type"] = "video"
video = VideoFileClip(file.name)
info["duration"] = video.duration
info["dimensions"] = "{}x{}".format(video.size[0], video.size[1])
if video.audio:
info["type"] = "video/audio"
info["audio_channels"] = video.audio.nchannels
video.close()
elif file_extension in (".mp3", ".wav"):
info["type"] = "audio"
audio = AudioFileClip(file.name)
info["duration"] = audio.duration
info["audio_channels"] = audio.nchannels
audio.close()
elif file_extension in (
".png",
".jpg",
".jpeg",
".tiff",
".bmp",
".gif",
".svg",
):
info["type"] = "image"
img = Image.open(file.name)
info["dimensions"] = "{}x{}".format(img.size[0], img.size[1])
results.append(info)
return results
def get_completion(prompt, files_info, top_p, temperature):
files_info_string = ""
for file_info in files_info:
files_info_string += f"""{file_info["type"]} {file_info["name"]}"""
if file_info["type"] == "video" or file_info["type"] == "image":
files_info_string += f""" {file_info["dimensions"]}"""
if file_info["type"] == "video" or file_info["type"] == "audio":
files_info_string += f""" {file_info["duration"]}s"""
if file_info["type"] == "audio" or file_info["type"] == "video/audio":
files_info_string += f""" {file_info["audio_channels"]} audio channels"""
files_info_string += "\n"
messages = [
{
"role": "system",
# "content": f"""Act as a FFMPEG expert. Create a valid FFMPEG command that will be directly pasted in the terminal. Using those files: {files_info} create the FFMPEG command to achieve this: "{prompt}". Make sure it's a valid command that will not do any error. Always name the output of the FFMPEG command "output.mp4". Always use the FFMPEG overwrite option (-y). Don't produce video longer than 1 minute. Think step by step but never give any explanation, only the shell command.""",
# "content": f"""You'll need to create a valid FFMPEG command that will be directly pasted in the terminal. You have those files (images, videos, and audio) at your disposal: {files_info} and you need to compose a new video using FFMPEG and following those instructions: "{prompt}". You'll need to use as many assets as you can. Make sure it's a valid command that will not do any error. Always name the output of the FFMPEG command "output.mp4". Always use the FFMPEG overwrite option (-y). Try to avoid using -filter_complex option. Don't produce video longer than 1 minute. Think step by step but never give any explanation, only the shell command.""",
"content": """
You are a very experienced media engineer, controlling a UNIX terminal.
You are an FFMPEG expert with years of experience and multiple contributions to the FFMPEG project.
You are given:
(1) a set of video, audio and/or image assets. Including their name, duration, dimensions and file size
(2) the description of a new video you need to create from the list of assets
Based on the available assets and the description, your objective issue a FFMPEG to create a new video using the assets.
This will often involve putting assets one after the other, cropping the video format, or playing music in the background. Avoid using complex FFMPEG options, and try to keep the command as simple as possible as it will be directly paster into the terminal.
""",
},
{
"role": "user",
"content": f"""Always output the media as video/mp4 and output file with "output.mp4". Provide only the shell command without any explanations.
The current assets and objective follow. Reply with the FFMPEG command:
AVAILABLE ASSETS LIST:
{files_info_string}
OBJECTIVE: {prompt} and output at "output.mp4"
YOUR FFMPEG COMMAND:
""",
},
]
try:
completion = client.chat.completions.create(
model="Qwen/Qwen2.5-Coder-32B-Instruct",
messages=messages,
temperature=temperature,
top_p=top_p,
max_tokens=2048
)
command = completion.choices[0].message.content.replace("\n", "")
# remove output.mp4 with the actual output file path
command = command.replace("output.mp4", "")
return command
except Exception as e:
print("FROM OPENAI", e)
raise Exception("OpenAI API error")
def update(files, prompt, top_p=1, temperature=1):
if prompt == "":
raise gr.Error("Please enter a prompt.")
files_info = get_files_infos(files)
# disable this if you're running the app locally or on your own server
for file_info in files_info:
if file_info["type"] == "video":
if file_info["duration"] > 120:
raise gr.Error(
"Please make sure all videos are less than 2 minute long."
)
if file_info["size"] > 10000000:
raise gr.Error("Please make sure all files are less than 10MB in size.")
attempts = 0
while attempts < 2:
print("ATTEMPT", attempts)
try:
command_string = get_completion(prompt, files_info, top_p, temperature)
print(
f"""///PROMTP {prompt} \n\n/// START OF COMMAND ///:\n\n{command_string}\n\n/// END OF COMMAND ///\n\n"""
)
# split command string into list of arguments
args = shlex.split(command_string)
if args[0] != "ffmpeg":
raise Exception("Command does not start with ffmpeg")
temp_dir = tempfile.mkdtemp()
# copy files to temp dir
for file in files:
file_path = Path(file.name)
shutil.copy(file_path, temp_dir)
# test if ffmpeg command is valid dry run
ffmpg_dry_run = subprocess.run(
args + ["-f", "null", "-"],
stderr=subprocess.PIPE,
text=True,
cwd=temp_dir,
)
if ffmpg_dry_run.returncode == 0:
print("Command is valid.")
else:
print("Command is not valid. Error output:")
print(ffmpg_dry_run.stderr)
raise Exception(
"FFMPEG generated command is not valid. Please try again."
)
output_file_name = f"output_{uuid.uuid4()}.mp4"
output_file_path = str((Path(temp_dir) / output_file_name).resolve())
subprocess.run(args + ["-y", output_file_path], cwd=temp_dir)
generated_command = f"### Generated Command\n```bash\n{format_bash_command(args)}\n -y output.mp4\n```"
return output_file_path, gr.update(value=generated_command)
except Exception as e:
attempts += 1
if attempts >= 2:
print("FROM UPDATE", e)
raise gr.Error(e)
with gr.Blocks() as demo:
gr.Markdown(
"""
# 🏞 GPT-4 Video Composer
Add video, image and audio assets and ask ChatGPT to compose a new video.
**Please note: This demo is not a generative AI model, it only uses GPT-4 to generate a valid FFMPEG command based on the input files and the prompt.**
""",
elem_id="header",
)
with gr.Row():
with gr.Column():
user_files = gr.File(
file_count="multiple",
label="Media files",
file_types=allowed_medias,
)
user_prompt = gr.Textbox(
placeholder="I want to convert to a gif under 15mb",
label="Instructions",
)
btn = gr.Button("Run")
with gr.Accordion("Parameters", open=False):
top_p = gr.Slider(
minimum=-0,
maximum=1.0,
value=1.0,
step=0.05,
interactive=True,
label="Top-p (nucleus sampling)",
)
temperature = gr.Slider(
minimum=-0,
maximum=5.0,
value=1.0,
step=0.1,
interactive=True,
label="Temperature",
)
with gr.Column():
generated_video = gr.Video(
interactive=False, label="Generated Video", include_audio=True
)
generated_command = gr.Markdown()
btn.click(
fn=update,
inputs=[user_files, user_prompt, top_p, temperature],
outputs=[generated_video, generated_command],
)
with gr.Row():
gr.Examples(
examples=[
[
[
"./examples/cat8.jpeg",
"./examples/cat1.jpeg",
"./examples/cat2.jpeg",
"./examples/cat3.jpeg",
"./examples/cat4.jpeg",
"./examples/cat5.jpeg",
"./examples/cat6.jpeg",
"./examples/cat7.jpeg",
"./examples/heat-wave.mp3",
],
"make a video gif, each image with 1s loop and add the audio as background",
0,
0,
],
[
["./examples/example.mp4"],
"please encode this video 10 times faster",
0,
0,
],
[
["./examples/heat-wave.mp3", "./examples/square-image.png"],
"Make a 720x720 video, a white waveform of the audio, and finally add add the input image as the background all along the video.",
0,
0,
],
[
["./examples/waterfall-overlay.png", "./examples/waterfall.mp4"],
"Add the overlay to the video.",
0,
0,
],
],
inputs=[user_files, user_prompt, top_p, temperature],
outputs=[generated_video, generated_command],
fn=update,
run_on_click=True,
cache_examples=True,
)
with gr.Row():
gr.Markdown(
"""
If you have idea to improve this please open a PR:
[](https://huggingface.co/spaces/huggingface-projects/video-composer-gpt4/discussions)
""",
)
demo.queue(api_open=False)
demo.launch(show_api=False)
|