Generating Graphic Novels With Deep Neural Networks — DALL-E Flow + GPT-3

Can a computer generate graphic novels?

Saverio Pulizzi
Better Programming

--

A collage with the 170 illustrations generated with Dalle-Flow from a GPT-3 text prompt.

Read this article to learn more about my journey on how I used deep neural networks to generate a sci-fi graphic novel entirely written and illustrated by a computer.

The most recent advancements in the field of deep learning have proven that computers are capable of generating creative content.

In this article, we will cover the following topics:

  • The Storyteller: GPT-3 ✍️
  • The Illustrator: DALLE-FLOW 🎨
  • The Publisher: Neural Canvas 🧠
  • The e-book: Infinite Creativity ∞

AI does not substitute human creativity but it empowers human senses to let them discover the unknown

The Storyteller: GPT-3 ✍️

The Generative Pre-trained Transformer 3 is an autoregressive language model that uses deep learning to produce human-like text.

It is the third-generation language prediction model in the GPT-n series created by OpenAI, a San Francisco-based artificial intelligence research laboratory.

GPT-3’s full version has a capacity of 175 billion machine learning parameters. The quality of the text generated by GPT-3 is so high that it can be difficult to determine whether or not it was written by a human.

We used this technology to generate a story starting from a few inputs.

OpenAI Playground Example

The Illustrator: DALLE·FLOW 🎨

DALL·E Flow is an interactive workflow for generating high-definition images from text prompt.

First, it leverages DALL·E-Mega to generate image candidates and then calls CLIP-as-service to rank the candidates w.r.t. the prompt.

The preferred candidate is fed to GLID-3 XL for diffusion, which often enriches the texture and background, and then upscaled to 1024x1024 via SwinIR.

DALL·E Flow is built with Jina in a client-server architecture, which gives it high scalability, non-blocking streaming, and a modern Pythonic interface.

For this project, we did not use DALLE·2 from OpenAI as we are still on their waiting list. However, we are happy with the results we got from DALL·E Flow so far!

An Infographic explaining how text to image models work compared to the human’s brain.

The Publisher: Neural Canvas🧠

Neural Canvas is the platform that we built to combine the power of GPT-3 and DALL·E Flow.

Our platform, through a simple interface, allows the end user to generate unique stories and their illustrations provided minimal text input, while automatically storing and protecting its Intellectual Property in the blockchain.

For example the Graphic NovelInfinite Creativity” was generated with just one sentence for each chapter!

Neural Canvas Landing Page

The Neural Canvas Illustrator will generate 100 illustrations that you will receive after few hours on your email address.

Create 100 AI Illustrations for $9.99

The platform is still in beta and we just released access to our demo.

Generate 100 Illustrations here.

The first e-book: Infinite Creativity ∞

Infinite Creativity has been written and illustrated in a few hours thanks to Neural Canvas.

The story, divided in 30 chapters, is about a humanoid robot going to space and trying to find solutions to major global issues, including the United Nations Sustainable Development Goals.

A snippet from Infinite Creativity

The author through Neural Canvas has first generated all the 170 illustrations and then automatically minted them in the blockchain.

A screenshot of the Rocket AI “Infinite Creativity” NFT Collection

This represents the way the author’s intellectual property is protected in the blockchain as well as being a web3 way of monetizing from their books.

Each NFT is unique and gives the buyer access to the full story.

Want to Connect?Hope you enjoyed this article and feel free to get in touch.

--

--