Telephone / app.py
mlmPenguin's picture
Create app.py
a636536 verified
import gradio as gr
import torch
from diffusers import FluxPipeline
from transformers import BlipProcessor, BlipForConditionalGeneration
# Set up device
device = "cuda" if torch.cuda.is_available() else "cpu"
# Load the FLUX.1-schnell text-to-image model via diffusers.
pipe = FluxPipeline.from_pretrained(
"black-forest-labs/FLUX.1-schnell",
torch_dtype=torch.bfloat16
)
pipe.enable_model_cpu_offload() # helps save VRAM
# Load an image captioning model (BLIP) to guess the prompt.
caption_processor = BlipProcessor.from_pretrained("Salesforce/blip-image-captioning-base")
caption_model = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-base")
caption_model.to(device)
def play_game(initial_prompt: str, rounds: int):
images = []
current_prompt = initial_prompt
# Loop for the number of rounds specified
for i in range(rounds):
# Generate an image with FLUX.1-schnell.
result = pipe(
current_prompt,
guidance_scale=0.0,
num_inference_steps=4, # adjust for speed vs. quality
generator=torch.Generator(device).manual_seed(42 + i)
)
img = result.images[0]
images.append(img)
# Use the captioning model to "guess" the prompt from the image.
inputs = caption_processor(images=img, return_tensors="pt").to(device)
output = caption_model.generate(**inputs)
guessed_prompt = caption_processor.decode(output[0], skip_special_tokens=True)
# Update current prompt with the guessed caption.
current_prompt = guessed_prompt
return images
# Build the Gradio interface.
demo = gr.Interface(
fn=play_game,
inputs=[
gr.Textbox(label="Initial Prompt", placeholder="Enter your starting prompt..."),
gr.Slider(minimum=1, maximum=10, step=1, label="Number of Rounds", value=3)
],
outputs=gr.Gallery(label="Generated Images"),
title="Flux Prompt Guessing Game",
description=(
"Enter an initial prompt and choose the number of rounds. "
"The game will generate an image using FLUX.1-schnell, then the AI "
"will guess the prompt from that image to generate the next one, and so on."
)
)
demo.launch()