Spaces:
Running
on
T4
Running
on
T4
Update description, fix font issues
Browse files- app.py +4 -3
- audiocraft/utils/extend.py +43 -7
app.py
CHANGED
|
@@ -150,7 +150,7 @@ def predict(model, text, melody, duration, dimension, topk, topp, temperature, c
|
|
| 150 |
output = output.detach().cpu().float()[0]
|
| 151 |
with NamedTemporaryFile("wb", suffix=".wav", delete=False) as file:
|
| 152 |
if include_settings:
|
| 153 |
-
video_description = f"{text}\n Duration: {str(initial_duration)} Dimension: {dimension}\n Top-k:{topk} Top-p:{topp}\n Randomness:{temperature}\n cfg:{cfg_coef} overlap: {overlap}\n Seed: {seed}"
|
| 154 |
background = add_settings_to_image(title, video_description, background_path=background, font=settings_font, font_color=settings_font_color)
|
| 155 |
audio_write(
|
| 156 |
file.name, output, MODEL.sample_rate, strategy="loudness",
|
|
@@ -171,11 +171,12 @@ def ui(**kwargs):
|
|
| 171 |
"""
|
| 172 |
with gr.Blocks(title="UnlimitedMusicGen", css=css) as demo:
|
| 173 |
gr.Markdown(
|
| 174 |
-
"""
|
| 175 |
-
# Disclaimer: This won't run on CPU only. Clone this App and run on GPU instance!!!
|
| 176 |
# UnlimitedMusicGen
|
| 177 |
This is your private demo for [UnlimitedMusicGen](https://github.com/Oncorporation/audiocraft), a simple and controllable model for music generation
|
| 178 |
presented at: ["Simple and Controllable Music Generation"](https://huggingface.co/papers/2306.05284)
|
|
|
|
|
|
|
| 179 |
"""
|
| 180 |
)
|
| 181 |
if IS_SHARED_SPACE:
|
|
|
|
| 150 |
output = output.detach().cpu().float()[0]
|
| 151 |
with NamedTemporaryFile("wb", suffix=".wav", delete=False) as file:
|
| 152 |
if include_settings:
|
| 153 |
+
video_description = f"{text}\n Duration: {str(initial_duration)} Dimension: {dimension}\n Top-k:{topk} Top-p:{topp}\n Randomness:{temperature}\n cfg:{cfg_coef} overlap: {overlap}\n Seed: {seed}\n Melody File:#todo"
|
| 154 |
background = add_settings_to_image(title, video_description, background_path=background, font=settings_font, font_color=settings_font_color)
|
| 155 |
audio_write(
|
| 156 |
file.name, output, MODEL.sample_rate, strategy="loudness",
|
|
|
|
| 171 |
"""
|
| 172 |
with gr.Blocks(title="UnlimitedMusicGen", css=css) as demo:
|
| 173 |
gr.Markdown(
|
| 174 |
+
"""
|
|
|
|
| 175 |
# UnlimitedMusicGen
|
| 176 |
This is your private demo for [UnlimitedMusicGen](https://github.com/Oncorporation/audiocraft), a simple and controllable model for music generation
|
| 177 |
presented at: ["Simple and Controllable Music Generation"](https://huggingface.co/papers/2306.05284)
|
| 178 |
+
|
| 179 |
+
Disclaimer: This won't run on CPU only. Clone this App and run on GPU instance!
|
| 180 |
"""
|
| 181 |
)
|
| 182 |
if IS_SHARED_SPACE:
|
audiocraft/utils/extend.py
CHANGED
|
@@ -7,6 +7,8 @@ import string
|
|
| 7 |
import tempfile
|
| 8 |
import os
|
| 9 |
import textwrap
|
|
|
|
|
|
|
| 10 |
from huggingface_hub import hf_hub_download
|
| 11 |
|
| 12 |
def separate_audio_segments(audio, segment_duration=30, overlap=1):
|
|
@@ -123,16 +125,50 @@ def hex_to_rgba(hex_color):
|
|
| 123 |
rgba = (255,255,0,255)
|
| 124 |
return rgba
|
| 125 |
|
| 126 |
-
def
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 127 |
try:
|
| 128 |
-
font = ImageFont.truetype(font_name,
|
| 129 |
-
except:
|
| 130 |
try:
|
| 131 |
-
font = ImageFont.truetype(
|
| 132 |
except:
|
| 133 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 134 |
return font
|
| 135 |
|
|
|
|
| 136 |
def add_settings_to_image(title: str = "title", description: str = "", width: int = 768, height: int = 512, background_path: str = "", font: str = "arial.ttf", font_color: str = "#ffffff"):
|
| 137 |
# Create a new RGBA image with the specified dimensions
|
| 138 |
image = Image.new("RGBA", (width, height), (255, 255, 255, 0))
|
|
@@ -149,7 +185,7 @@ def add_settings_to_image(title: str = "title", description: str = "", width: in
|
|
| 149 |
text_x = width // 2
|
| 150 |
text_y = height // 2
|
| 151 |
# Draw the title text at the center top
|
| 152 |
-
title_font =
|
| 153 |
|
| 154 |
title_text = '\n'.join(textwrap.wrap(title, width // 12))
|
| 155 |
title_x, title_y, title_text_width, title_text_height = title_font.getbbox(title_text)
|
|
@@ -158,7 +194,7 @@ def add_settings_to_image(title: str = "title", description: str = "", width: in
|
|
| 158 |
title_draw = ImageDraw.Draw(image)
|
| 159 |
title_draw.multiline_text((title_x, title_y), title, fill=font_color, font=title_font, align="center")
|
| 160 |
# Draw the description text two lines below the title
|
| 161 |
-
description_font =
|
| 162 |
description_text = '\n'.join(textwrap.wrap(description, width // 12))
|
| 163 |
description_x, description_y, description_text_width, description_text_height = description_font.getbbox(description_text)
|
| 164 |
description_x = max(text_x - (description_text_width // 2), description_x, 0)
|
|
|
|
| 7 |
import tempfile
|
| 8 |
import os
|
| 9 |
import textwrap
|
| 10 |
+
import requests
|
| 11 |
+
from io import BytesIO
|
| 12 |
from huggingface_hub import hf_hub_download
|
| 13 |
|
| 14 |
def separate_audio_segments(audio, segment_duration=30, overlap=1):
|
|
|
|
| 125 |
rgba = (255,255,0,255)
|
| 126 |
return rgba
|
| 127 |
|
| 128 |
+
def load_font(font_name, font_size=16):
|
| 129 |
+
"""
|
| 130 |
+
Load a font using the provided font name and font size.
|
| 131 |
+
|
| 132 |
+
Parameters:
|
| 133 |
+
font_name (str): The name of the font to load. Can be a font name recognized by the system, a URL to download the font file,
|
| 134 |
+
a local file path, or a Hugging Face model hub identifier.
|
| 135 |
+
font_size (int, optional): The size of the font. Default is 16.
|
| 136 |
+
|
| 137 |
+
Returns:
|
| 138 |
+
ImageFont.FreeTypeFont: The loaded font object.
|
| 139 |
+
|
| 140 |
+
Notes:
|
| 141 |
+
This function attempts to load the font using various methods until a suitable font is found. If the provided font_name
|
| 142 |
+
cannot be loaded, it falls back to a default font.
|
| 143 |
+
|
| 144 |
+
The font_name can be one of the following:
|
| 145 |
+
- A font name recognized by the system, which can be loaded using ImageFont.truetype.
|
| 146 |
+
- A URL pointing to the font file, which is downloaded using requests and then loaded using ImageFont.truetype.
|
| 147 |
+
- A local file path to the font file, which is loaded using ImageFont.truetype.
|
| 148 |
+
- A Hugging Face model hub identifier, which downloads the font file from the Hugging Face model hub using hf_hub_download
|
| 149 |
+
and then loads it using ImageFont.truetype.
|
| 150 |
+
|
| 151 |
+
Example:
|
| 152 |
+
font = load_font("Arial.ttf", font_size=20)
|
| 153 |
+
"""
|
| 154 |
+
|
| 155 |
try:
|
| 156 |
+
font = ImageFont.truetype(font_name, font_size)
|
| 157 |
+
except (FileNotFoundError, OSError):
|
| 158 |
try:
|
| 159 |
+
font = ImageFont.truetype(font_name, font_size)
|
| 160 |
except:
|
| 161 |
+
try:
|
| 162 |
+
req = requests.get(font_name)
|
| 163 |
+
font = ImageFont.truetype(BytesIO(req.content), font_size)
|
| 164 |
+
except:
|
| 165 |
+
try:
|
| 166 |
+
font = ImageFont.truetype(hf_hub_download(".assets", font_name), encoding="UTF-8")
|
| 167 |
+
except:
|
| 168 |
+
font = ImageFont.load_default()
|
| 169 |
return font
|
| 170 |
|
| 171 |
+
|
| 172 |
def add_settings_to_image(title: str = "title", description: str = "", width: int = 768, height: int = 512, background_path: str = "", font: str = "arial.ttf", font_color: str = "#ffffff"):
|
| 173 |
# Create a new RGBA image with the specified dimensions
|
| 174 |
image = Image.new("RGBA", (width, height), (255, 255, 255, 0))
|
|
|
|
| 185 |
text_x = width // 2
|
| 186 |
text_y = height // 2
|
| 187 |
# Draw the title text at the center top
|
| 188 |
+
title_font = load_font(font, 26) # Replace with your desired font and size
|
| 189 |
|
| 190 |
title_text = '\n'.join(textwrap.wrap(title, width // 12))
|
| 191 |
title_x, title_y, title_text_width, title_text_height = title_font.getbbox(title_text)
|
|
|
|
| 194 |
title_draw = ImageDraw.Draw(image)
|
| 195 |
title_draw.multiline_text((title_x, title_y), title, fill=font_color, font=title_font, align="center")
|
| 196 |
# Draw the description text two lines below the title
|
| 197 |
+
description_font = load_font(font, 16) # Replace with your desired font and size
|
| 198 |
description_text = '\n'.join(textwrap.wrap(description, width // 12))
|
| 199 |
description_x, description_y, description_text_width, description_text_height = description_font.getbbox(description_text)
|
| 200 |
description_x = max(text_x - (description_text_width // 2), description_x, 0)
|