DrawGPT - Use Only GPT to Draw Any Image!

[kind monolith]

Use only GPT to draw any image! No diffusers! 🤖 🎨 This app uses only GPT-3 DaVinci to draw images... no DALL-E or anything! 🚀

LINK: https://drawgpt.ai
TWITTER: https://twitter.com/DrawGPT
TWITTER BOT: https://drawgpt.ai/twitter-bot/

**How does it do it !? 🧐 **

Javascript + HTML Canvas! 🎉 Images are rendered live at 512px x 512px and can be saved as PNG and used in any other art program including DALL-E. What about the code? See the Javascript used for your image so you can re-render it at any time in any program that supports Javascript.

**Share & Save Your Images 👥 💾 🙂 **

You can copy the link to any image you draw and share it or even set it as your PFP on Twitter. Login in with your Twitter or Web 3 wallet to start saving them to your account!

The images are goofy & fun. Check out the website, draw some images, and see what other people have made. Thank you for the support! 🙏 🤖 🎨

https://drawgpt.ai

Created by:
@kind monolith

[pine crest]

This looks so cool. Looking forward to checking it out soon

[simple lynx]

Its a bit buggy RN but im also looking forward to it aswell. Cant access the DrawGPT part.

[copper sparrow]

Cool

[kind monolith]

what do you mean that you cannot access the DrawGPT part? https://drawgpt.ai/

[kind monolith]

thanks!

[sick glacier]

POS images genter

[safe ice]

@kind monolith how can gpt draw it?

With a code generator?

Used in ChatGPT

[placid crescent]

Draw

[kind monolith]

what do you mean?

[kind monolith]

I use the GPT-3 DaVinci API (basically the same LLM that runs ChatGPT) & have it create images using code, you can see the Javascript code it uses to draw any image on the website when it generates them 🙂

@safe ice here is a portrait I just had it draw of you, if you click through on the link you can see the code it used in the Javascript Code section to draw these using an HTML canvas 2D canvas context: https://drawgpt.ai/prompt/4194/

@safe ice here is a blog post I wrote about it that is more targeted towards AI researchers & how to get their LLMs to be visually aware and visually enabled, but it also covers some of how I got GPT to draw: https://kickiniteasy.com/drawgpt-make-ai-art-draw-images-using-an-ai-that-only-knows-text/

[kind monolith]

it works as a Twitter reply bot, not a Discord bot unfortunately

[safe ice]

I thought it was generating svg :)

[kind monolith]

SVG is too verbose for output with the tokens, it won't always get the tags closed, I tried SVG originally but it just didn't work well