In 2022, we saw text-to-image models like DALL-E2, Midjourney, and Stable Diffusion and conversational bots like ChatGPT. Then we saw many alternatives that were built on different languages. Now you can generate images, chat like ChatGPT, and more on a single Discord server. 

Developed by Kaveen Kumarasinghe (Kav-K), a computer engineering student at University of Waterloo, the server, called GPT3E, offers a robust, but simple interface of GPT-3 based chatbot along with a text-to-image art generator using DALL-E 2.

Click here to check it out!

With the latest update of GPT3E, users can chat on Discord with –  !g <prompt> and have endless conversations with it as the model can generate summarization of the conversation using GPT-3. You can also edit previously entered prompts.

For image generation, users can input their prompt using –  !draw <prompt> and also generate multiple images with different characteristics and qualities. The bot also optimises the given prompt to be suitable for DALL-E based image generation. 

The project uses Python and openai-async rewrite for generation of threads. Developers can also add the bot to their server following the steps given in the GitHub repository. 

Other Chatbots and Image Generators

Now developers can also build ChatGPT-like chatbots using the open-source alternative to GPT-3, Google’s PaLM. Phil Wang posted code on GitHub where developers can build chatbots using RLHF (Reinforcement Learning with Human Feedback) on Python.

Moreover, Since Stable Diffusion got open-source we have seen many new innovations by the AI art enthusiasts. One was a macOS native app, Amazing AI, for text-to-image generation, Riffusion for music generation using Stable Diffusion, and Unstable Diffusion, an NSFW AI art generator.