Speech recognition
ASR that actually hears Tigrigna.
General-purpose Whisper handles Tigrigna at low quality and with no awareness of dialect or domain. Tigvoice is fine-tuned on a curated Tigrigna corpus — conversational audio, broadcast clips, and reviewed WhatsApp voice notes — with chunked alignment and human-validated gold sets.
- Word-level timestamps with confidence scores
- Streaming for live captions; batch for archives
- Punctuation and Geez-script normalization built in
- Diarization-ready output for multi-speaker recordings
API preview
POST /v1/asr
# curl example curl https://api.tabotec.ai/v1/asr \ -H "Authorization: Bearer $TABOTEC_KEY" \ -F "[email protected]" \ -F "language=tig" \ -F "timestamps=word" # response { "text": "ሰላማት፣ ስምካ መን ይበሃል?", "segments": [ { "start": 0.42, "end": 2.81, "text": "ሰላማት፣ ስምካ መን ይበሃል?", "confidence": 0.94 } ] }
API preview
POST /v1/tts
# curl example curl https://api.tabotec.ai/v1/tts \ -H "Authorization: Bearer $TABOTEC_KEY" \ -H "Content-Type: application/json" \ -d '{ "text": "እንኳዕ ብደሓን መጻእኩም።", "voice": "tig-female-warm", "format": "wav" }' --output greeting.wav
Speech synthesis
Voices that sound like people, not robots.
Tigvoice TTS produces natural Tigrigna speech with several voice presets — male, female, formal, conversational. Built for IVR, accessibility, broadcast subtitling read-out, and language-learning tools.
- Multiple voice IDs with consistent identity across calls
- SSML-style controls for pacing, pauses, and emphasis
- 16 kHz and 24 kHz output, WAV / MP3 / OGG
- Custom voice cloning available with consented speaker data
Built for real use cases
Where Tigvoice goes first.
Broadcast subtitling
Eritrean and Tigrayan radio and television have decades of unsubtitled archives. Tigvoice ASR turns those archives searchable and translatable.
Telecom IVR
Voice menus, balance checks, and customer service in Tigrigna — read by Tigvoice TTS, recognized by Tigvoice ASR.
Healthcare & gov services
Hotlines and digital agents that meet citizens in their first language. We work with Ministry-of-Health-grade pilot frameworks.
Diaspora apps
Voice notes that get transcribed for accessibility, language learning, search, and search-in-place across community archives.
NGO programs
Field interviews and community feedback get transcribed in-country, on local hardware, with no audio leaving the network.
Education
Mother-tongue learning content authored once, voiced consistently, captioned automatically — all in Tigrigna.
Early access
Tigvoice is in pilot.
We're onboarding a small group of pilot users in late 2026. Join the waitlist and tell us what you'd use it for.