Sawti: AI Voice Studio for Tunisian Darija
Project type: SaaS platform & API
The project
Sawti is an AI content studio for Arabic, French and English, with speech-to-text, text-to-speech and LLMs built for Tunisian and Algerian darija.
Most AI tools only speak Modern Standard Arabic. Sawti understands and speaks Tunisian and Algerian darija, so content sounds local rather than translated. Teams use it to transcribe, voice, visualize and animate their marketing content, and developers can use its models through an API.
Key features
- Speech-to-Text: transcription of Arabic (including Tunisian and Algerian) and English audio, from an uploaded file or a recording
- Text-to-Speech: audio generation with a choice of female and male voices
- Adjustable voice precision and expressiveness
- Image Studio and Video Studio for AI-generated visuals and videos
- API and API key management for developers
- Job queue with history, copy and download of results
- Credits, billing and analytics dashboard
- Arabic, French and English interface
Artificial intelligence
Darija speech recognition
Speech-to-text models adapted to Tunisian and Algerian dialects, including sentences that mix Arabic and French.
Natural speech synthesis
Text-to-speech models that produce natural Tunisian voices, with adjustable temperature and expressiveness.
Tunisian dialect LLM
A language model that understands and generates content in Tunisian dialect.
Screenshots
