Generate Pokémon images from text prompts using Stable Diffusion with an interactive Gradio interface.
This project is a quick demo built in about 1 hour to showcase text-to-image generation with Stable Diffusion. It uses the pre-trained "runwayml/stable-diffusion-v1-5" model and adds a Gradio UI for real-time interaction. Originally, I aimed to fine-tune on the "diffusers/pokemon-gpt4-captions" dataset, but time constraints led to a polished base-model demo instead—fine-tuning is next on the list!
- Text-to-Image: Enter prompts like "A cool Pokémon in anime style" to generate unique images.
- Gradio Interface: Interactive web UI for easy use, shareable via a public link.
- Built Fast: From idea to GitHub in about 1 hour, using Colab and local tweaks.
- Stable Diffusion: "runwayml/stable-diffusion-v1-5" (Hugging Face Diffusers)
- Gradio: For the interactive frontend
- PyTorch: Backend for GPU-accelerated generation
- Python: Core scripting in Jupyter Notebook
-
Clone the Repo: git clone https://github.com/anakintano/StableDiffusion_pokemon.git cd StableDiffusion_pokemon
-
Install Dependencies:
pip install diffusers torch transformers accelerate gradio
-
Run the Notebook:
- Open
StableDiffusion_Pokemon.ipynbin Jupyter:jupyter notebook StableDiffusion_Pokemon.ipynb
- Execute all cells to launch the Gradio interface.
- Use the local URL (e.g.,
http://127.0.0.1:7860) or the public link (ifshare=True) to generate images.
- Open
-
Try a Prompt:
- Example: "A cool Pokémon in anime style"
Here’s a sample generated image:
- What Worked: Base model generation and Gradio UI were seamless.
- What Didn’t: Fine-tuning hit dataset access issues and time limits—future iterations will tackle this.
- Next Steps: Fine-tune on a Pokémon dataset, add multi-image generation, and optimize performance.
Feel free to fork, tweak, or suggest improvements! Open an issue or PR—I’m here to collaborate.
MIT License—see LICENSE if added, or assume open-source for now.
Built by anakintano | March 2025