Dollyglot

Executive Summary

Dollyglot is a YC W25 seed-stage company building a consumer AI companion app where users upload a photo and voice to create a real-time video avatar "AI friend" — best described as Character.AI with real-time video. The team is technically credible: they co-built Parler-TTS Mini Multilingual (open-source, Hugging Face-published) and claim the world's first photo-to-real-time video avatar model, both verifiable signals of genuine frontier AI capability. The traction story, however, is thin: the only monetization metric on record is 1,000 paying users for a pre-Dollyglot predecessor product, and the actual Dollyglot platform had not yet launched publicly as of its YC listing. The single biggest risk is a convergence of regulatory exposure and competitive dynamics — the product processes biometric facial data, explicitly targets Gen Alpha (children), and competes in a space already attracting congressional hearings and regulatory action, all while being a 2-person team with $500K raised against incumbents like Character.AI ($150M+ raised, 20M MAU) and Meta AI (effectively free distribution at planetary scale).