Talking Spell: A Wearable System Enabling Real-Time Anthropomorphic Voice Interaction with Everyd...
summary
A wearable system called Talking Spell turns any object into a talking persona by combining an eye‑level camera (OV2640 on an ESP32‑S3 scope) and a wand trigger; it uses YOLO (object recognition), QWEN‑VL for persona generation, Whisper for speech‑to‑text and TTS for voice output. Interaction follows acquaintance, familiarization, bonding stages. Technical tests on eight objects train new items in under 10 minutes and report high recognition accuracy; a 12‑participant user study shows positive engagement across entertainment, companionship, creativity, and utility.
AI-written summary · about
chapters (10)
- 0:00Introduction and motivation
- 1:08Talking Spell system overview
- 2:36Hardware and interaction stages
- 4:00Acquaintance and familiarization
- 5:25Bonding and real-time interaction
- 6:07Use cases and applications
- 7:52Evaluation and user study
- 8:58Future improvements and conclusion
- 9:19Q&A session part 1
- 11:26Q&A session part 2
related talks
20mMake an Amazing World pt.2
37mPersonality Is the New Ringtone: Experiments in AI design
23mExperimenting with Hardware and AI to Create Dynamic Live Streaming User Experiences
20mBuilding the next interface - AI personas who feel as natural as interacting with a human
02mIntuitive Brain-Computer Interface Control Using Onomatopoeia for an Enhanced Gaming Experience
28mWhat’s Needed in Tech to bring Voice to the Future
33mDesigning Voice Agents for Real Conversations
43mWearable Live Captions