Inspiration
Planning a route to explore a new/ unfamiliar place can often be daunting for tourists, as there are many factors to consider, for example, which locations to visit, the optimal route to do so. In addition, it is likely that a tourist would walk to know a bit of the background of the location they are visiting, and often audio guides require a certain price. Therefore, we are inspired to create an app that does the route planning for the tourist, and provides a cheaper way in accessing audio guides for new places.
What it does
It is a PWA that allows the user to log in with their Google account. After the user has log in, it accesses the user's current location and asks a few questions to figure out their preferences, for example the distance they are willing to walk, their budget etc. It then generates an optimised route based on their response. When the user goes to one of the locations, it provides an audio guide to inform the user about the significance of the location, a transcript will also be displayed whilst the audio is playing.
How we built it
Front-end: Vue and Tailwind (Nuxt) Back-end: Python (FastAPI) Database: MongoDB APIs: Google Maps, ElevenLabs (TTS), Gemini AI (transcript generation for podcast)
Challenges we ran into
Trying to find the optimal route. We wanted to include at least one location from each category that the user might be interested in (for example, museums, galleries). However, we also need to factor in the distance the user is willing to walk, therefore, some problem solving was required to generate a route that includes at least one location from one of the categories whilst being close/ in the range of the desired walking distance.
In addition, one of our team members, Keanu, spent 6 arduous hours wrestling with a Python library named podcastfy, an open source alternative to NotebookLM which uses Gemini to generate transcripts and ElevenLabs to generate voice lines to be combined into a "audio guide" podcast. Initially, the library did not support providing both an audio file and a word-by-word accurately timestamped transcript, so Keanu had to painstakingly modify the podcastfy library function-by-function in order to support this.
Accomplishments that we're proud of
Getting a working prototype and including all of the core features.
What we learned
A lot :D
What's next for WalkieTalkie
Allow user to rate their trip and in upcoming trips, give user recommendations based on previous ratings
Log in or sign up for Devpost to join the conversation.