Giving my wife a voice with Ruby and AI cloning (+ discussion)
summary
Built a Ruby backend that sends text to 11 Labs and returns binary audio to Twilio to place calls; a one‑minute original clip produces a recognisable clone and a recorded demo call works, and a cloned voice even defeats a phone‑based voice authenticator. Also shows vision and retrieval examples using Claude 3 / Anthropic for QC, vector embeddings for RAG, and a fine‑tuned model that cuts manual data‑entry from two people to about 25% remaining.
AI-written summary · about
chapters (16)
- 0:00Introduction
- 3:48AI voice cloning for Peggy
- 6:15Live demo phone call
- 8:01Dark side of AI cloning
- 10:50Ruby AI app architecture
- 12:15AI API for Ruby devs
- 12:57Vision API garment demo
- 15:04Neural networks explained
- 19:01AI prompting techniques
- 23:15Knowledge retrieval with AI
- 30:38Vector embeddings search
- 32:46Fine-tuning AI models
- 35:16Local AI with open source
- 36:39Mini AI call center demo
- 41:00Future of AI in Ruby
- 42:24Q&A
related talks
25mConversation with Sean King (SVP, GM at Veritone) and @ianutile
30mSteal this talk
17mMy AI Clone Attended Conferences While I Flew Planes - Tanisha Sharma
51mBuilding an AI assistant that makes phone calls
58mClone yourself with Azure Custom Neural Voice
01h 02m[Full Workshop] Building Conversational AI Agents
52mGoing Full Duplex - Building Apps for Speech
39m"Steal This Talk"