Credits

Lingo Kaiava's spoken voices and speech recognition are built on open-source work. Thank you to everyone below.

Engines

• Kokoro text-to-speech (Apache 2.0)
• Piper text-to-speech (MIT)
• Supertonic 3 text-to-speech, our main voice for Korean, Turkish and most European and Middle-Eastern languages (code MIT; model weights under the BigScience OpenRAIL-M license, whose use restrictions apply)
• whisper.cpp / OpenAI Whisper speech recognition (MIT)

Piper voices

• Mexican Spanish "claude" (high) — Apache 2.0
• European Spanish "davefx" — CC0
• Argentine Spanish "daniela" (high) — CC BY-SA 4.0
• German "thorsten" (high) — CC0 (Thorsten Müller)
• Polish "bass" (high) — Apache 2.0
• Ukrainian "mykyta" (high) — Apache 2.0
• Swedish "alma" — CC BY 4.0, trained on the NST Swedish speech data (Språkbanken, National Library of Norway)
• Vietnamese "vais1000" — CC BY 4.0 (VAIS-1000 corpus)
• Greek "rapunzelina", Czech "jirka", Hungarian "anna", Romanian "mihai", Finnish "harri", Danish and Norwegian "talesyntese", Dutch "nathalie", Russian "denis" — CC0
• Arabic "kareem" and Indonesian "news_tts" — community voices published through the Piper voice collection

Voice collection: rhasspy/piper-voices.

Questions or corrections: contact us.