# Copyright (2024) Bytedance Ltd. and/or its affiliates # # This source code is licensed under the license found in the # LICENSE file in the root directory of this source tree. from __future__ import annotations import os import random import uuid import gradio as gr import numpy as np from loguru import logger from caller import ( SeedT2ICaller, SeedEditCaller ) from PIL import Image help_text = """ ## How to use this Demo Step 1. Type in the caption/instruction text box, and click "Generate" to generate an initial image using Seed-T2I. Step 2. Type in the caption/instruction text box, and click "Edit" to edit the current image using SeedEdit. This is a demo with limited QPS and a simple interface. For a better experience, please use [Doubao](https://www.doubao.com/chat/)/[Dreamina](https://dreamina.capcut.com/ai-tool/image/generate) APP. - The current demo does not support multi-round editing, which may lead to overexposure with multiple rounds of upload and download edits. - Higher-quality input images will produce higher-quality edited results. For low-quality images, unwanted changes, e.g. facial id, may occur. Note: This demo is governed by the license of CC BY-NC 4.0 \ We strongly advise users not to knowingly generate or allow others to knowingly generate harmful content, \ including hate speech, violence, pornography, deception, etc. \ (注:本演示受CC BY-NC的许可协议限制。我们强烈建议,用户不应传播及不应允许他人传播以下内容,\ 包括但不限于仇恨言论、暴力、色情、欺诈相关的有害信息。) """ example_instructions = [ "Make it a picasso painting", "close its eye", "convert to a bronze statue", "make it wearing a hat", "make it wearing a PhD suit", "Turn it into an anime.", "have it look like a graphic novel", "make it gain weight", "what would he look like bald?", "Have it smile", "Put in a cocktail party.", "move to the beach.", "add dramatic lighting", "Convert to black and white", "What if it were snowing?", "Give a leather jacket", "Turn into a cyborg!", ] def main(): resolution = 1024 cfg = {"resolution": resolution} model_t2i = SeedT2ICaller(cfg) cfg_edit = {} model_edit = SeedEditCaller(cfg_edit) logger.info("All models loaded") def load_example(): example_image = Image.open(f"uni_test/test.jpg").convert("RGB") example_instruction = random.choice(example_instructions) edited_image, example_instruction = generate(example_image, example_instruction, cfg_scale=0.5) return example_image, example_instruction, edited_image def generate_t2i(instruction: str, cfg_scale: float = 0.5): if not instruction: return None, "" logger.info("Generate images ...") # Call model and capture the status gen_image, success = model_t2i.generate(instruction, batch_size=1, cfg_scale=cfg_scale) if not success or gen_image is None: logger.error("Image generation failed or returned None. please retry") return None, instruction return gen_image, instruction def generate(input_image: Image.Image, instruction: str = None, cfg_scale: float = 0.5): logger.info("Generating images ...") if not instruction or input_image is None: return input_image, "" logger.info("Running diffusion models ...") edited_image, success = model_edit.edit(input_image, instruction, batch_size=1, cfg_scale=cfg_scale) if not success or edited_image is None: logger.error("Image editting failed or returned None.") return None, instruction return edited_image, instruction def reset(): return None, None, "" with gr.Blocks(css="footer {visibility: hidden}") as demo: with gr.Row(): with gr.Column(scale=1, min_width=100): generate_button = gr.Button("Generate") with gr.Column(scale=1, min_width=100): edit_button = gr.Button("Edit") with gr.Column(scale=1, min_width=100): load_button = gr.Button("Load Example") with gr.Column(scale=1, min_width=100): reset_button = gr.Button("Reset") with gr.Row(): with gr.Column(scale=3): instruction = gr.Textbox(lines=1, label="Caption/Edit Instruction", interactive=True, value=None) with gr.Column(scale=1): cfg_scale = gr.Slider(value=0.5, minimum=0.0, maximum=1.0, step=0.1, label="Edit/Text Strength (CFG)", interactive=True) with gr.Row(): input_image = gr.Image(label="Input Image", type="pil", interactive=False, height=resolution, width=resolution) edited_image = gr.Image(label="Edited Image", type="pil", interactive=False, height=resolution, width=resolution) gr.Markdown(help_text) load_button.click( fn=load_example, inputs=[], outputs=[input_image, instruction, edited_image] ) generate_button.click( fn=generate_t2i, inputs=[instruction, cfg_scale], outputs=[input_image, instruction] ) edit_button.click( fn=generate, inputs=[input_image, instruction, cfg_scale], outputs=[edited_image, instruction] ) reset_button.click( fn=reset, inputs=[], outputs=[input_image, edited_image, instruction] ) # demo.launch(server_name="0.0.0.0", server_port=8024) demo.queue().launch(share=False) if __name__ == "__main__": main()