QlooCast

Inspiration

We set out to explore how we could use Qloo’s cultural intelligence in an effective and innovative way. After brainstorming several ideas, we settled on podcast recommendations. This led to the development of QlooCast, a solution for people who want to listen to more podcasts than time allows by delivering personalized recommendations and summarized digests.

What it does

QlooCast allows users to share their interests and then provides them with suggested podcasts as well as a concise summary in both text and audio formats. It combines Qloo’s cultural intelligence with AI transcription and summarization, giving users the opportunity to discover podcasts they may not have known about and receive summaries that save time—helping them focus only on full episodes that truly interest them:

• Gather Interests: Users share their tastes (movies, music, books, cuisine, travel, and more).
• Qloo Recommendations: We send this data to Qloo’s Insights API, which returns a list of podcasts they are likely to enjoy.
• Top Podcast Selection: OpenAI analyzes the results and ranks the podcasts provided by Qloo.
• Transcription: The top five podcasts are sent for transcription by Whisper.
• Summaries: OpenAI takes the transcriptions and creates a podcast digest for the user. Each selected podcast also includes a concise summary plus an explanation of why it was chosen.
• Audio Digest: The final personalized digest is also delivered as audio, allowing users to consume it however they prefer.

How we built it

• Qloo’s Insights recommends podcasts based on user inputs and how they align with their tastes.
• OpenAI ranks and finalizes the podcast list.
  
• Whisper transcribes the top ranked podcast episodes.
  
• OpenAI summarizes the transcribed podcast episodes into a podcast digest.
  
• OpenAI uses text-to-speech to generate an audio version of the podcast digest.
  

Challenges we ran into

• Qloo’s API: Our main challenge was understanding and mapping Qloo’s API parameters to reliably retrieve the most relevant podcast recommendations based on the user's input.
• Response time: We needed to take into account server timeout limits for Vercel’s free plan when doing transcriptions.

What's next for QlooCast

• User Accounts: Enable user accounts and weekly digests, allowing people to continuously receive personalized podcast summaries without re-entering their preferences.
• Exclude Older Podcasts: Remove all podcasts from the recommendations that are older than a month.

Built With

• openai
• qloo
• react