Highlights
- Natural Voice and Accent Tuning
Human-like speech quality was achieved by combining multiple TTS models, fine-tuning parameters, and incorporating authentic Chilean Spanish voice samples. - Rapid Delivery
The complete PoC was built in approximately 38 engineering hours, demonstrating DataArt’s ability to rapidly prototype sophisticated multimodal AI solutions. - Creative Avatar Engineering from Limited Assets
Using only a static bottle image, DataArt engineered a workaround pipeline to generate a lifelike, animated avatar suitable for live interaction, demonstrating strong problem-solving in generative media. - Platform-Agnostic, Production-Oriented Design
The PoC architecture supports expansion into additional hospitality and retail scenarios, including virtual concierges and enhanced digital experiences powered by multilingual conversational AI.
Technologies Used
Avatar Creation:
- HeyGen
- D-ID
Voice Models:
- ElevenLabs
- ElevenLabs V3
- Fish
- Starfish
- Panda
