Cutting edge speech to text
at your fingertips.
Accurate transcripts and AI insights from every conversation, in the languages Africa speaks. Fully multilingual, so nothing gets lost when a call switches language mid-sentence.
- languages in production, with code-switching
- 22+
- African countries with live models
- 11
- hours of audio transcribed every month
- ~900k
Your data is hosted and processed in the EU (Ireland) and deleted 24 hours after upload.
Languages across Africa
Built for the way Africa speaks.
Fully multilingual.
Production speech models in 11 countries, from isiZulu and Afrikaans to Kiswahili, Amharic, Twi and Wolof. They follow conversations that switch language mid-sentence, too.
- Live speech models
- Publicly available
- Supported, not yet publicly released
- Languages with production speech models
- 22+
- African countries with live speech models
- 11
Choose a country
South Africa
11 languages- English
- isiZulu
- isiXhosa
- Afrikaans
- Sepedi
- Setswana
- Sesotho
- Xitsonga
- siSwati
- Tshivenda
- isiNdebele
Used for contact centres, Parliament, broadcast news and lectures
Available now
All 11 South African languages, plus Kiswahili
- English“Hello”
- isiZulu“Sawubona”
- isiXhosa“Molo”
- Afrikaans“Hallo”
- Sepedi“Thobela”
- Setswana“Dumela”
- Sesotho“Lumela”
- Xitsonga“Avuxeni”
- siSwati“Sawubona”
- Tshivenda“Ndaa”
- isiNdebele“Lotjhani”
- Kiswahili“Habari”
Code-switching, built in
“Yebo, ngifuna uku-check i-balance yami, because i-statement angiyitholanga this month.”
isiZuluEnglish, in the same sentence
Also supported
Not available for public release yet
- Amharic
- French
- Kalenjin
- Kikuyu
- Luganda
- Luo
- Portuguese
- Somali
- Twi
- Wolof
Our models support these languages too, but they aren’t publicly released.
Need one of them? Talk to usConversation intelligence
Don’t just search your audio.
Understand it.
Our speech AI turns calls, meetings and broadcasts into insight you can act on: summaries, sentiment, quality scores and live alerts, in every language your customers speak.
Audio
Calls, meetings and broadcasts
Transcript
In every language spoken
Intelligence
Summaries, sentiment, scores and alerts
Contact-centre analytics are delivered with Callbi, powered by Saigen. Talk to us about your calls
Proud partner
The speech engine behind 
Callbi’s call-centre speech analytics run on Saigen speech recognition. Every call is transcribed, summarised and scored, across South Africa’s mixed-language conversations.
Visit callbi.io(opens in a new tab)- businesses and contact centres analyse their calls with Callbi
- 150+
Services
Speech recognition for every kind of conversation
From contact centres to newsrooms, Saigen turns hours of audio into accurate transcripts, AI insights and analytics you can act on.
Contact centres
Speech analytics for every call: AI summaries, sentiment, quality scores and compliance tracking.
Media monitoring
Follow radio and TV broadcasts, with keyword alerts the moment a name or topic is mentioned.
Transcriptions
Customised, affordable and highly accurate transcripts you can review in our self-editing interface.
Personal information redaction
Protect sensitive information by automatically redacting private data from your transcripts.
Custom words & custom models
Tuned to your words and your domain
General-purpose speech recognition gets close. A model built for your domain gets the names, products and jargon your people actually say.
Custom words
Give us the names, places, brands and terms that matter to you. We add them to your model, so they are transcribed correctly and can trigger keyword alerts the moment they’re said.
- Saigen, before the keywordMissed
…sitting next to me sherwin salary k abuja…
- General-purpose modelMissed
…sitting next to me shehu salawu aboki…
- Saigen, keyword addedDetected
…sitting next to me sherwin Solanke-Ebhojie abuja…
Custom models
Specialists beat generalists. A model tuned to your domain is as good as a general-purpose system, or better, at a fraction of the cost. The secret is the data: targeted, domain-specific corpora, collected or curated by a team with nearly two decades of experience.
- Contact centres14 languages
- Broadcast news14 languages · 7 countries
- Parliament11 languages
- University lecturesEnglish
Got a language, accent or vocabulary nobody handles well? That’s exactly what we build. Talk to us about a custom model
About Saigen
Decades of speech research, now in production across Africa
Saigen was founded in 2017, as an Alphawave company, to take decades of speech research into production. Our founders built the first speech recognition for South Africa’s official languages. Today our models run in 11 African countries.
- Prof Etienne BarnardCo-founder, Non-executive Director
- Dr Charl van HeerdenCo-founder, Executive Director: Research and AI
Where we come from
CSIR
The first speech recognition for all 11 official South African languages.
NWU · IARPA Babel
Part of the winning consortium, building spoken term detection for 26 under-resourced languages.
Google
Worked with Google’s speech team on Voice Search for Afrikaans, South African English and isiZulu.
Saigen · 2017 to today
Custom speech recognition, analytics and voices in production for contact centres, broadcasters, parliament and universities.
Trusted by teams at
Turn any audio into intelligence
Start with an accurate transcript today, or talk to us about AI insights across all your audio.
Contact us
We'd love to hear from you
Talk to the team
Whether you want to automate customer service or work with your recordings more easily, we are always happy to hear from you and answer your questions.