I’ve built a few prototypes with the OpenAI voice to text API with code largely written using Cursor. This has been fast and easy to incorporate into Next.js apps. I can add an audio-recording-to-text feature to any app in a couple of minutes, ready for use in a production environment.
There are several other options for voice to text as well. MacOS has a built-in voice to text feature and there are several other Whisper wrappers available, some that can run locally. Talking is much faster than typing and allows me to capture raw thoughts faster, which I can then refine later. LLMs are also quite good at structuring these raw thoughts into a more refined form that I can then edit.
This is not to replace writing, but actually to make it easier for me to write more frequently. There are lots of times when I have an idea that I want to capture and return to later. The more I can capture when it’s top of mind, the more likely I can come back later and do something with it.