text-decorator / app.py
christinac's picture
Update app.py
c8e1566
raw history blame
No virus
2.36 kB
import os
import shutil
import torch
import gradio as gr
MY_SECRET_TOKEN=os.environ.get('HF_TOKEN_SD')
from PIL import Image,ImageFont,ImageDraw
from gradio.mix import Series
#from io import BytesIO
from diffusers import StableDiffusionImg2ImgPipeline
YOUR_TOKEN=MY_SECRET_TOKEN
device="cuda" if torch.cuda.is_available() else "cpu"
pipe = StableDiffusionImg2ImgPipeline.from_pretrained("runwayml/stable-diffusion-v1-5", use_auth_token=YOUR_TOKEN)
pipe.to(device)
#draw an image based off of user's text input
def drawImage(text, text_size, prompt, strength, guidance_scale): #(text, text_size, font)
out = Image.new("RGB", (512, 512), (0, 0, 0))
#add some code here to move font to font-directory
font = './font-directory/DimpleSans-Regular.otf'
fnt = ImageFont.truetype(font, int(text_size))
d = ImageDraw.Draw(out)
d.multiline_text((16, 64), text, font=fnt, fill=(255, 255, 255))
#init_image = out
out.save('initImage.png')
images = []
images = pipe(prompt=prompt, init_image=out, strength=strength, guidance_scale=guidance_scale).images
#images[0].save = ("image.png")
#images = []
#images.append(out)
#out.show()
return images[0]
#def newImage(image, prompt):
#return images
#drawImage = gr.Interface(fn=drawImage, inputs=gr.Textbox(placeholder="shift + enter for new line",label="what do you want to say?"),outputs="image")
#newImage = gr.Interface(fn=newImage,inputs=[gr.Textbox(placeholder="prompt",label="how does your message look and feel?")],outputs="image")
#demo = gr.Series(drawImage,newImage)
demo = gr.Interface(
title="this will be very slow since it's on CPU! msg me if you would like to try",
#description="christina",
fn=drawImage,
inputs=[
gr.Textbox(placeholder="shift + enter for new line",label="what do you want to say?"),
#"file"
gr.Number(label="text size",value=240),
gr.Textbox(placeholder="eg. imagery, art style, materials, emotions",label="how does your message look and feel?"), #figure out models in series
gr.Slider(label="strength (how much noise will be added to the input image)",minimum=0, maximum=1, step=0.01, value=0.7),
gr.Slider(label="guidance scale (how much the image generation follows the prompt)",value=15, maximum=20),
],
outputs="image")
demo.launch()