A Remake of the Google Gemini Fake Demo, Except Using GPT-4 and It's Real

Original video here.

Heads-to-heads comparison of Gemini Pro and GPT-4.

how to build

• clone this repo, cd into it
• duplicate .env.example and name the copy .env
• fill out the VITE_OPENAI_KEY= value with your OpenAI api key. you must have access to the gpt-4-vision-preview model
  • you can also try out the Gemini API if you have a key -- fill out VITE_GEMINI_KEY in the same .env
• then, run:
• npm install
• npm run dev
• the demo will be running at http://localhost:5173

note: the in-browser speech recognition works best in Google Chrome

TODO

• allow input of API keys as <input> on the page
• deploy frontend to site i.e. sagittarius.greg.technology via vite+github actions
• enable streaming output..!
• make new video with 1) uses of repo in the wild / forks 2) UI improvements 3) streaming output / comparison
• enable selection of dictation language
• add allcontributors bot
• add dependabot