Academic research
Research-grade Unity lab: ReadyPlayer.me avatars, 11Labs voice cloning & Firebase configs in one month
Purdue University Virtual Lab
- Delivery
- ~1 month
- Participant scale
- 5,000+
- Team
- 1 dev + PM
- Status
- Active research
The problem
Researchers required highly customizable avatars (including consistent branded garments on user-created RPM bodies), AI voices that could clone participant demographics from brief samples, real-time behavioural feedback tied to task performance, and scalable experiment configuration — with no off-the-shelf path for MP3-based TTS inside Unity at runtime or RPM-level asset replacement.
How we approached it
- 01 Discovery with Purdue on experimental design, RPM constraints and voice pipeline requirements
- 02 Integrated ReadyPlayer.me with custom asset authoring and runtime mesh-level garment replacement
- 03 Implemented 11Labs TTS/voice cloning with a custom native MP3→WAV converter for Unity playback
- 04 Added Firebase Firestore for centralized experiment parameters and dynamic player feedback hooks
What we shipped
An updated Unity research build with RPM avatar pipeline, Firestore-driven configuration, 11Labs-backed character dialogue with cloned voices, custom audio conversion so streamed speech plays reliably in-session, and integrated positive/negative feedback tied to participant behaviour — plus dependency modernization and documented code for hand-off.
Helped us a lot, both with our actual project and then again when we encountered some hiccups with Upwork. We immediately plan to rehire them.
Planning something in this space?
Jump to the contact form on our site to start a conversation.