Inspiration
Vocabulary retention is critical for language learning. Research shows words stick in long-term memory through repeated exposure in meaningful contexts. Stories provide this naturally. I built StoryOdyssey to generate unlimited personalized stories around target vocabulary. Each story exposes learners to keywords multiple times in different contexts. This creates stronger memory associations than isolated flashcards. AI makes it possible to create tailored content at scale for effective vocabulary acquisition.
What it does
StoryOdyssey generates personalized stories from keywords in 11 languages. Users hover over words for instant translations, save vocabulary, and review with flashcards for repeated exposure and retention.
How we built it
Built with Flask backend and React frontend. Used TheBloke's Chronomaid-Storytelling-13B quantized model (GGUF format) for story generation, Google Translator handles translations, and Stanza/Fugashi parse all sentences for interactive word learning and word translation.
Challenges we ran into
I couldn't run the story generator model on my GPU efficiently, so the story generating process was taking about 3-5 minutes on my machine. I used the GGUF quantized version to reduce memory requirements.
Accomplishments that we're proud of
Created a seamless workflow from keyword input to interactive story with instant translations. Successfully integrated 11 languages with proper tokenization and parsing.
What's next for StoryOdyssey
Adding text-to-speech (TTS) for pronunciation practice and audio learning to enhance the immersive language learning experience.
Log in or sign up for Devpost to join the conversation.