Sophon-Realtime
by Adventists AI · Model · Voice & audio
unproven
Only Adventists AI's own claims so far; nobody has measured or reviewed it.
Full-duplex voice model that detects tone, emotion and word stress; zero-shot voice cloning; three variants (flash, expressive with meta tags, end-to-end); all Apache 2
AccessOpen weightsLicenseApache 2.0 (maker's statement; confirm it covers all three variants)
Why we say so
The evidence is one description from the maker: full-duplex voice, tone and emotion detection, zero-shot voice cloning, and three variants. No benchmark, tester or community report backs any of it. Apache 2.0 is stated by the maker with no conflicting source.