Promptmakr

Introduction
Other App screens
Gemini AI automatic conversion of prompt

Inspiration

AI image generation is becoming super popular and more and more tools are popping up like Midjourney, Dall-E, Gemini, Co-Pilot, Stable Diffusion etc. all allowing users to generate stunning images using text prompts. What we realized that it takes a lot of trial and error to get the correct effect and not everyone is familiar with the nuances of writing good prompts for these image generation tools. This we believe has opened up opportunities for AI image prompt artists to sell their prompts by turning their prompts into dynamic prompts with variables that can be tweaked by the buyer generate veriations but still able to achieve the style they purchased. Currently there is no platform that allows them to do that automatically. PROMPTMAKR provides this platform for AI image artists.

What it does

PROMPTMAKR allows AI image artists to enter their prompt text and with a single press of a button turn it into a dynamic variable based prompt. In addition to that, the tool also generates description of the variables it identifies and writes instructions on how to use these variables while allowing the user to edit or modify it.

How we built it

The tool was built using a no/low code app called FlutterFlow as we to validate the market and understand the demand for configurable prompts for AI Image and text engines. Initially we had built this app where users would write everything manually. So we updated it to now use Gemini AI to take the load away from the user and the AI takes the input form the end user and does everything automatically. We trained the model based on a collection of keyboards and prompts that are verified. We protect our API keys by calling cloud Firebase functions. This protects not only our keys but also allows us to make small tweaks to the calls to Gemini API without having to update the app.

Challenges we ran into

We have built our app using. Flutterflow (low code tool). Flutterflow does allow Gemini API to be integrated, but It only works with the default Gemini model. For our usecase, we needed to make use of a custom model we had trained on Google AI Studio. We use Google AI studio and Firebase cloud functions to call the Gemini model. It would be great if we can specify a trained model of Gemini to use instead of passing training data in every call.

Accomplishments that we're proud of

The biggest accomplishment for us is to make the prompt creation process super quick and simple. Our novice users were struggling to understand how to create prompts with variables and hot to design instructions for using the variables. Gemini helped us automate this entire process with a single click of a button.

What we learned

We learned that using the Google AI Studio's code generation makes calling Gemini API much easier than we had expected it to be. It would be awesome if it dumped out Dart code so we can call it directly from the app but using Cloud provides security for the key so for the short term it is an acceptable compromise.

What's next for Promptmakr

Next we want to add payment services so that people can sell their prompts and in turn making this truly into a bustling marketplace. We want to make it easier to using standard AI image generation engines to automatically load the images for the prompt as well as integrate non-image based prompt in the near future.