TL;DR: AI interviews you, on whatever topics you want, creating audio archives and transcripts.
Work in progress, not ready to try, developing in public, check back later 😎
- I want an AI Interviewer, as a way to motivate myself to talk about a variety of topics that I'd like to share with the world.
- Experiment with AI in the browser via WASM.
- Contribute back to open source
- SvelteKit - fullstack TS front + backend
- Deploy to Cloudflare Pages (at minimum)
- Audio recording via browser
- Audio transcription via
- WASM LLM in browser or
- LLM via API, eg OpenAI
- Next question generation via:
- WASM LLM in browser or
- LLM via API, eg OpenAI
- Save audio files to Cloudflare R2 (at minimum)
- TTS (Text to Speech) via openAI in browser
- Works on cloudflare pages
- record user audio via browser
- STT user audio
- Whisper WASM in browser
- whisper API
- Browser Speech API?
- on chrome is sent to google servers
- behind flag on Firefox
- feed convo to llm to determine next question
- prefer: WASM llm - mistral7b or better
- alt: gpt4[-turbo] api
- save audio (ai and human) to cf r2
- stitch together chunks in player to play back interview
- download full interview as mp3