import discord from discord.ui import Button import gradio_client from gradio_client import Client import gradio as gr import os import threading #for deepfloydif import requests import json import random from PIL import Image import matplotlib.pyplot as plt import matplotlib.image as mpimg import time # random + small llama # #todos #alert #fix error on first command on bot startup #stable diffusion upscale #buttons for deepfloydIF (1,2,3,4) #application commands instead of message content checks (more user-friendly) DFIF_TOKEN = os.getenv('DFIF_TOKEN') #deepfloydIF #df = Client("DeepFloyd/IF", DFIF_TOKEN) #not reliable at the moment df = Client("huggingface-projects/IF", DFIF_TOKEN) #stable diffusion upscaler sdlu = Client("huggingface-projects/stable-diffusion-latent-upscaler", DFIF_TOKEN) # Set up discord bot class MyClient(discord.Client): async def on_ready(self): print('Logged on as', self.user) async def on_message(self, message): #safety checks---------------------------------------------------------------------------------------------------- # tldr, bot should run if #1) it does not have @offline role #2) user has @verified role #3) bot is in #bot-test channel # bot won't respond to itself, prevents feedback loop + API spam if message.author == self.user: return # if the bot has this role, it won't run OFFLINE_ROLE_ID = 1103676632667017266 # 1103676632667017266 = @offline / under maintenance guild = message.guild bot_member = guild.get_member(self.user.id) if any(role.id == OFFLINE_ROLE_ID for role in bot_member.roles): return # the message author needs this role in order to use the bot REQUIRED_ROLE_ID = 897376942817419265 # 900063512829755413 = @verified, 897376942817419265 = @huggingfolks if not any(role.id == REQUIRED_ROLE_ID for role in message.author.roles): return # channels where bot will accept commands ALLOWED_CHANNEL_IDS = [1100458786826747945] # 1100458786826747945 = #bot-test if message.channel.id not in ALLOWED_CHANNEL_IDS: return #deepfloydif---------------------------------------------------------------------------------------------------- if message.content.startswith('!deepfloydif'): # change to application commands, more intuitive #(prompt, negative_prompt, seed, number_of_images, guidance_scale,custom_timesteps_1, number_of_inference_steps, api_name="/generate64") #-> (stage_1_results, stage_1_param_path, stage_1_result_path) # input prompt prompt = message.content[12:].strip() negative_prompt = '' seed = 0 number_of_images = 4 guidance_scale = 7 custom_timesteps_1 = 'smart50' number_of_inference_steps = 50 stage_1_results, stage_1_param_path, stage_1_result_path = df.predict( prompt, negative_prompt, seed, number_of_images, guidance_scale, custom_timesteps_1, number_of_inference_steps, api_name='/generate64') #stage_1_results, stage_1_param_path, stage_1_result_path = df.predict("gradio written on a wall", "blur", 1,1,7.0, 'smart100',50, api_name="/generate64") # stage_1_results -> path to directory with png files, so we isolate those png_files = [f for f in os.listdir(stage_1_results) if f.endswith('.png')] # merge images into larger, 2x2 image the way midjourney does it if png_files: first_png = png_files[0] second_png = png_files[1] third_png = png_files[2] fourth_png = png_files[3] ''' [],[],[],[] -> [][] [][] ''' first_png_path = os.path.join(stage_1_results, first_png) second_png_path = os.path.join(stage_1_results, second_png) third_png_path = os.path.join(stage_1_results, third_png) fourth_png_path = os.path.join(stage_1_results, fourth_png) img1 = Image.open(first_png_path) img2 = Image.open(second_png_path) img3 = Image.open(third_png_path) img4 = Image.open(fourth_png_path) # create a new blank image with the size of the combined images (2x2) combined_image = Image.new('RGB', (img1.width * 2, img1.height * 2)) # paste the individual images into the combined image combined_image.paste(img1, (0, 0)) combined_image.paste(img2, (img1.width, 0)) combined_image.paste(img3, (0, img1.height)) combined_image.paste(img4, (img1.width, img1.height)) # save the combined image combined_image_path = os.path.join(stage_1_results, 'combined_image.png') combined_image.save(combined_image_path) embed = discord.Embed(title="Combined Image") embed.set_image(url="attachment://combined_image.png") components = discord.ui.MessageComponents() components.add_action_row( Button(style=ButtonStyle.blue, label="Image 1", custom_id="image_1"), Button(style=ButtonStyle.blue, label="Image 2", custom_id="image_2"), Button(style=ButtonStyle.blue, label="Image 3", custom_id="image_3"), Button(style=ButtonStyle.blue, label="Image 4", custom_id="image_4"), ) await message.channel.send(embed=embed, file=discord.File(combined_image_path, 'combined_image.png'), components=components) async def on_button_click(self, interaction): if interaction.custom_id.startswith("image_"): image_index = int(interaction.custom_id.split("_")[1]) await interaction.respond(content=f"You clicked image {image_index}") #dfif stage 2 selected_index_for_stage_2 = image_index - 1 seed_2 = 0 guidance_scale_2 = 4 custom_timesteps_2 = 'smart50' number_of_inference_steps_2 = 50 result_path = df.predict(stage_1_result_path, selected_index_for_stage_2, seed_2, guidance_scale_2, custom_timesteps_2, number_of_inference_steps_2, api_name='/upscale256') # Send the upscaled image in a new message upscale_image_path = os.path.join(result_path, "output.png") with open(upscale_image_path, 'rb') as f: await interaction.channel.send('Here is the upscaled image:', file=discord.File(f, 'upscaled_image.png')) DISCORD_TOKEN = os.environ.get("GRADIOTEST_TOKEN", None) intents = discord.Intents.default() intents.message_content = True client = MyClient(intents=intents) def run_bot(): client.run(DISCORD_TOKEN) threading.Thread(target=run_bot).start() def greet(name): return "Hello " + name + "!" demo = gr.Interface(fn=greet, inputs="text", outputs="text") demo.launch()