Five months back, I started building Visioncast. Drawing inspiration from psychology-related books like "Psychocybernetics" and "What to Say When You Talk to Yourself", I've been fascinated by the power of affirmations and positive self-talk. This fascination was the seed for Visioncast, guiding users through personalized audio exercises including affirmations, meditations, motivational speeches, and visualizations, all tailored to their individual needs.<p>I'm leveraging GPT-4 and TTS to deliver content that's not just personalized but deeply resonant with the challenges and aspirations of the user. Whether it's bolstering self-esteem or stopping procrastination, the app is aims to support mental and emotional well-being.<p>Perfecting the user experience has been tricky. Right now I'm copying ChatGPT's suggestion cards to start you out, paired with a customization screen that offers extensive control over the final output. This level of personalization sets Visioncast from app like Calm, Headspace, and Waking Up, which offer no personalization.<p>In the first version, it took 1m 20s to generate audio. I had to wait for the whole GPT4 response to come in. I've streamlined audio processing and streaming (with lots of fighting with AWS/Django/React Native and Nginx) And now, with the app inching ever closer to what I envisioned as an AI mentor, the validation from our first two paying users has been incredibly affirming.<p>I'd love to hear your insights/feedback, as I'm still working out UX and some bugs.