Neurokõne vs TGSpeechBox

Tartu Ülikooli keeletehnoloogide arendatud eestikeelne kõnesüntees. Valida saab 10 erineva kõneleja vahel ning kõne kiirust on võimalik reguleerida. Meie hääli saab kasutada ka tekstelemendi lugejana, valides seadetes neist ühe vaikimisi eestikeelseks hääleks. Süntesaatoriks on kasutatud tehisnärvivõrkudel põhinevat mudelit, mis on treenitud eestikeelsetel uudiste ja ilukirjanduse korpustel. Meie kõnesüntees on saadaval ka veebilehel neurokone.ee --- Estonian text-to-speech synthesis developed by the NLP research group at the University of Tartu. There are 10 speakers to choose from and the speech tempo can be adjusted. Any voice can also be used in Speak Selection when set as the default Estonian voice on the device. Our engine is based on a neural network trained on Estonian news and literature corpora. Our text-to-speech service is also available at neurokone.ee

TGSpeechBox is a high-quality speech synthesizer that works as a system voice with VoiceOver and as a standalone text-to-speech app. Features: • 26 languages with natural-sounding speech • Full VoiceOver integration — use as your primary system voice • 7 voice presets plus YAML-defined voice profiles • Deep engine tuning: pitch mode, voice tone, formant sharpness, inflection, tremor, breathiness, and more • Phoneme editor: fine-tune individual speech sounds with live preview • Pronunciation dictionary: add custom word pronunciations with IPA override support • Stress, compound, and character dictionaries for advanced control • Export and share your customizations • Adjustable sample rate (11025–44100 Hz) • Rate override and rate boost for VoiceOver • Language lock: keep your preferred language regardless of system setting • Runs entirely on-device — no internet required, no data leaves your phone TGSpeechBox uses a Klatt-style formant synthesis engine with eSpeak-ng for phoneme generation. It is open source under the MIT license. Built by a blind developer, designed for accessibility from day one.