Clean, readable, and responsive code generation, reinforced with GPT-4 Vision to achieve production-ready code.
Start your 14-day free trial now! Notable features include semantic element detection, auto fixing designs and intelligent Tailwind rounding.
Hey fellow hackers! π€
We've been developing our Figma to code plugin, π₯FireJet, lovingly for the past two years.
The reason why we started in the first place is because we found that while there are several figma to code plugins already out there, the code quality that they produced was never up to par.
The three aspects that we've been religiously focusing on over the past two years were:
1. Code readability
2. Pixel Perfection
3. Responsiveness
To achieve the above, we've added two major features - GPT 4 vision and Engine 2.
π€ GPT 4 Vision:
OpenAI has recently launched their vision API, which is able to provide more context for better code generation.
Transformer models have always been extremely capable at producing nice to read code - and it's something we've always wanted to add into FireJet. The problem is that when it edits the code, it often messes up the design.
However, now with the Vision API, we are able to make modifications to the generated code to make it cleaner, more readable and responsive, which is why we've added it as a feature to augment the code output from FireJet.
β‘Engine 2:
However, even with GPT 4 vision we reached a ceiling when it came to code quality, which is the same ceiling every other figma to code plugin faced - a garbage in, garbage out situation.
A lot of the time, Figma designs are very nice, but when put through Figma to code plugins, they would produce horribly nested, absolute positioned elements
This is because most Figma to code converters rely on the structure of the user's figma design. The problem with that when creating figma designs, the focus is on speed and beauty, rather than on perfect nesting for outputting to code
To overcome this problem, we developed a proprietary algorithm based on chess engines, which we believe separates us far for the rest in terms of code quality. (You can read more about it at https://www.firejet.io/blog/What...)
(Okay but feel free to flame me at philip@firejet.io if it doesn't work perfectly)
Hope you guys like it! Let us know if you have any feedback or comments!
Report
Congrats to Stephen and Philip! Amazing product!! π
In the future when A.I. owns the entire development cycle nobody will care if the code is maintainable so things like Tailwind support won't be needed.
@kingromstar Some day in the future! For now, let's all enjoy the time saved with low-code tools π
Report
Incredible journey, team FireJet! π The dedication to code quality, readability, and responsiveness, coupled with the integration of GPT 4 Vision and Engine 2, truly sets FireJet apart. Excited to see this revolution in Figma to code plugins! π₯
Wow! @stephenalvin97, your FireJet is amazing! This Figma to Tailwind Workflow augmented with GPT-4 Vision looks like it's going to make a big splash in the market π Congratulations on the launch, and I wish you all the success in your endeavors!
@shankaikai So engine two is able to produce a largely pixel perfect output - but it lacks things like whether it should be a button, form or something else.
What GPT 4 vision is able to do is add those additional buttons, plus sometimes its able to refactor the code to make it easier to read.
The reason why it being GPT4 + vision is important is because without the vision context, GPT tends to mess up the original design instead of improve it
Replies
FireJet
FireJet
Licode (Now Appaca)
FireJet
Go Mail Merge
FireJet
FireJet
FireJet
FireJet
FireJet
FireJet
FireJet