github kizuna-ai-lab/sokuji v0.44.0

5 hours ago

Highlight: hear the other side, or sit beside them

Face-to-face (Soniox and KizunaAI). In Both mode, set Other side to Beside me: two people share one microphone, each wears one earbud, and each hears only the translation into their own language. (#613, #624)

Translation I hear. The other side's translation can now be read aloud to you in a voice of its own, on Soniox, KizunaAI, Google Gemini, OpenAI Realtime, OpenAI Live, Palabra AI, Doubao AST 2.0 and Free. The audio settings are now arranged by who hears what, each row with its own output device and channel; Text Only is now the Translation the other side hears switch. (#615, #618, #621)

Improvements

  • Speaker labels (Soniox): once a second person speaks, the other side's lines are numbered in the conversation, the subtitle window, the LAN caption page and exports. (#613)
  • Echo cancellation now removes everything the computer plays on Windows and macOS, so a meeting app's audio from the speakers is not translated as your voice. (#614)
  • Windows: when the microphone will not open, Sokuji names the app holding it. (#623)
  • Wording: the language section says "I hear" when the other side's translation is read to you. (#625)

Fixes

  • Edge TTS voices (Free) speak again after Microsoft moved its read-aloud service; every earlier version is affected. The extension asks for one new permission for this, and Chrome pauses it until you accept. (#622)
  • Settings links open their section again after the panel was closed quickly. (#616)

Full Changelog: v0.43.4...v0.44.0

Don't miss a new sokuji release

NewReleases is sending notifications on new releases.