Hi folks. I wrote a bot that generates a podcast that reports the weather forecast for Brooklyn, NY. I’m using OpenAI’s /completions endpoint to add some editorial summarization of the weather forecast I pull from the National Weather Service.
10 Likes
Cool idea, and love that you open-sourced it. The weather report itself is quite short though (~30s), are you considering expanding the amount of content that’s generated? Love the concept of weather and a joke, I think it’d be interesting to give the outputs more personality with finetuned models and voice cloning (ie respeecher, resemble.ai, overdub).
3 Likes
Hey @asabet. Thanks for taking a look! The forecast is as long or as short as the NWS provides in their “detailed forecast” for the day and the night of the episode. I decided to keep it concise instead of tacking on forecasts for future days.
I definitely want to evolve the personality of the AI part, and will look at fine-tuned models when I do.
Voice cloning sound very cool, but I don’t want to spend significant money on this little art project. Current operating cost is ~$2.50/month, and about $1 of that is Amazon Polly. It looks like the solutions you recommended cost a lot more.
1 Like
I don’t want to spend significant money. Your weather podcast bot for Brooklyn, NY sounds like a fantastic project! Leveraging OpenAI’s /completions endpoint to enhance the editorial summarization of the National Weather Service forecast adds an insightful touch.
Connecting this with the concept of a snowday predictor Canada, it’s exciting to see technology being harnessed to provide timely and relevant weather information. Such initiatives not only serve local communities but also showcase the creative potential of AI in addressing everyday needs.
Looking forward to learning more about the how and why behind your innovative bot.