Podmate App

Podmate is a mobile app that allows you to effectively “talk back” to podcasts. With both text and voice mode functionality, you can ask questions, explore a topic deeper and fact check something in real time using AI.

Podmate automatically transcribes your voice, and passes it into an AI model, along with a transcript of a podcast for a quick and interactive podcast listening experience

The idea for Podmate came to me when i was driving. I was listening to a podcast and i was so engaged with the conversation, i wanted to find out more about what the speakers were saying and verify the facts.

However it’s inconvenient and dangerous to do that in normal places where you listen to podcasts, such as in the gym or when you’re driving.

With my previous experience with the react ecosystem, I have decided to use React Native to develop the app, as it was the fastest and most familiar to set up. I began by creating a simple Expo App, and developed each screen sequentially, utilizing Firebase for authentication, database management, and cloud functions.

I’ve learned a lot building server side functions using firebase functions as i needed to create communication between the app, to the openai api’s for AI services, and the google speech api for voice transcription.

Billing was also a challenge to implement. As I had not worked with any app that implemented in-app billing before, i had to go deep into learning how ios handles in app purchases, and connect that all with firebase for smooth purchases.

It should be noted that i had extensively used AI to build this app using cursor. Tying in this new workflow, i managed to save time looking at documentation, and hence managed to make a functional minimum viable product in less than 2 weeks.

At the moment, the app is in its initial launch stages, with only a few beta users. Reviews have been positive so far, and feedback has been positive.

The next step is to ramp up advertising and marketing campaigns, and to track analytics and ensure that users have an easy time submitting feedback, or reporting bugs in the app if anything has gone wrong. Additional news articles and publicity is always a positive.

The biggest problem with real-time communication with AI at the moment is long response times, sometimes more than a few second, representing a significant challenge with AI technology. In the future there are ways to mitigate this and increase the speed of responses. One solution is to use the last 10 or 20 seconds snippet of a podcast, from the time of the question, ensuring that the entire podcast transcript won’t need to be loaded upon starting the chat.

Other ways to improve latency is to use different AI providers such as Groq, using specialised chips for faster computation.