Daily Vietnamese is an autonomously posting LLM generated content application for Vietnamese language learning practice. It leverages generative LLMs to create conversation topics and dialogs, and uses GCP cloud functions, text-to-speech API, Pub/Sub, and cloud storage to asynchronously process the creation of conversation audio files and images based on a given topic. It publishes a new conversation each day.
- Typechat for topic and dialog generation
- OpenAI for image generation based on the conversation topic
- GCP text-to-speech AI for Vietnamese language audio generation
- GCP Cloud Functions for serverless execution
- GCP Pub/Sub for asynchronous processing
- GCP Cloud Scheduler for scheduling the creation of new topics and posting a new conversation each day
- GCP Cloud Storage for audio and image storage
- Algolia for searching and filtering content
- Sendgrid for emailing the daily conversation
- Prisma, Postgres, and CockroachDB for modeling and storing application data
- Next.js client using server components for rendering the content
- Tailwind CSS for styling
- Vercel for hosting the client application
- Autonomously creates conversation topics
- Generates Vietnamese language audio and a related image based on the conversation topic
- Publishes a new conversation each day
- Users can search and filter indexed content
- Users can listen to audio files of each conversation
- Users can listen to audio files of each individual word in a conversation
- Users can subscribe to get conversations emailed to them daily
- Add speech tonal analysis for practicing pronunciation of Vietnamese tones. Give user visual feedback based on adherence to the tone curve as compared to the reference audio files
- Add a flashcard feature for practicing vocabulary from the conversations
- Track user progress and suggest flashcards based on the user's performance
- Use spaced repetition algorithms