This is a fun demo of a full-stack ML app. It takes your text prompt as input and uses three models to produce four sample Pokémon card images — the models are StableDiffusion, an RNN for Pokémon name generation, and a basic OpenCV background removal model.
This app is an example of what can be done these days by what I'm calling model stacking, for lack of a better name. It's become so easy to stick together ML models, often without training many or all of them yourself.
demo link: modal-labs-example-text-to-pokemon-fastapi-app.modal.run/
cloud platform: modal.com
The code is here: github.com/modal-labs/modal-examples/tree/main/06_gpu_and_ml/text-to-pokemon
(Be aware that in the video the prompts used are previously seen and cached. Unseen prompt generations take 30-120 seconds)
Top comments (0)