We build in the open

Explore the research behind our models. We are committed to transparency and community-driven development in West African AI.

paper

Oolel-Embed: Cross-lingual Speech-Text Retrieval for Wolof and French

Oolel-Embed maps spoken Wolof queries and written French documents into a shared embedding space, enabling direct cross-lingual retrieval without a separate transcription-and-translation chain.

Yaya Sy, Dioula Doucouré, Christophe Cerisara, Irina Illina
2026-04-20
SpeechEmbeddingsInformation RetrievalWolofFrenchMatryoshka
Read More
engineering

Oolel Voices: Open-Source Text-to-Speech for Wolof

Oolel Voices is our open-source text-to-speech model for Wolof and mixed Wolof-French. Try the interactive demo to generate natural speech with voice cloning.

Soynade Research
2026-03-12
Text-to-SpeechWolofVoice CloningOpen Source
Read More
Demo
article

Soynade's Open Source Month

Announcing Soynade's Open Source Month: four weeks of models, datasets, and tools for African languages.

Yaya Sy, Dioula Doucouré
2026-02-18
Open SourceSynthetic DatavLLMNLPWolofBambara
Read More
article

Soynade Research joins UNICEF Venture Fund's Data and Trust Cohort

Soynade Research joins UNICEF Venture Fund as one of 7 organizations selected from 1,000 applications to build open-source AI solutions for West African languages.

Research Team
2026
UNICEFVenture FundOpen SourcePartnership
Read More
paper

Speech Language Models for Under-Represented Languages: Insights from Wolof

We present the first speech language model for Wolof, achieving 18.4% ASR improvement through continued pretraining on 860 hours of spontaneous speech and integration with a Wolof LLM.

Yaya Sy, Dioula Doucouré, Christophe Cerisara, Irina Illina
2025-09-25
SpeechLLMHuBERTWolofASR
Read More