The STORM models.
One family of models for speaking, listening and translating in African languages. In the app and the API you choose a voice and a language; we route each request to the model that does it best.
| Model | Languages | Speak | Clone a voice | Listen | Translate | Stream |
|---|---|---|---|---|---|---|
| STORM TTS | Yoruba, Hausa, Igbo, Nigerian Pidgin and Nigerian English | – | – | – | ||
| STORM Ensemble 3.0 | 64 languages across Africa | – | – | – | – | |
| STORM-OS V3 | 70+ languages across Africa | – | – | – | ||
| STORM ASR 2.0 | Yoruba, Hausa, Igbo, Nigerian Pidgin and Nigerian English, and speech that mixes them | – | – | – | ||
| STORM MT | English, Yoruba, Hausa, Igbo and Nigerian Pidgin | – | – | – | – |
STORM TTS
Text to speechPOST /v1/speechOur speech model for Nigerian languages. Native speakers, tested with native listeners, read your text the way people say it: naira amounts, phone numbers, local names and sentences that switch language halfway.
- Native voices for every core language
- Streams audio as it is generated
- Reads money, numbers and dates naturally
- Handles code-switched sentences
STORM Ensemble 3.0
Text to speechPOST /v1/speechOne speech model for the continent. Each language gets voices of its own, at studio quality, so a product can speak to Nairobi, Accra and Johannesburg as naturally as it speaks to Lagos.
- 166 named voices
- Studio-grade 24 kHz audio
- Voices native to each language
STORM-OS V3
Text to speechPOST /v1/speechThe widest reach in the family. It speaks languages from Wolof to Amharic, and carries a cloned voice across them, so one brand voice can be heard in every market.
- 70+ African languages
- Cloned voices across languages
- One model for the whole catalogue
STORM ASR 2.0
Speech to textPOST /v1/transcriptionsTranscribes recorded audio and live streams in the five core languages and speech that mixes them, with no language setting. In our benchmark it made the fewest mistakes of every model tested, in every language.
- No language setting needed
- Code-mixed speech
- Real-time streaming
- Timed segments for captions
STORM MT
TranslationPOST /v1/translationsTranslates text between English, Yoruba, Hausa, Igbo and Nigerian Pidgin, with tone marks kept, and hands it straight to STORM to speak.
- Tone marks kept
- Hands off to speech in one pipeline
Hear the models on your own words.
Request access to the PlotWeaver app, or bring us one customer journey and we will show you what STORM sounds like on it.