Transcription apps for Sesotho: Lesotho speech

Lesedi FM, the SABC's Sesotho station: transcription apps compared for Sesotho, in Lesotho, South Africa and Zimbabwe.

This page compares transcription apps and services for Sesotho.

Speakers. About 5.6 million native speakers (Ethnologue 18, 2015) plus about 7.9 million second-language speakers in South Africa (2002). The 2011 South African census recorded almost four million first-language speakers.

Writing. Latin (the Sesotho alphabet). Two written standards exist, a Lesotho orthography and a South African orthography. Neither marks tone.

Where Sesotho is recorded

  • Radio. Lesedi FM, the SABC's Sesotho station, launched in 1960 and broadcasts from Bloemfontein; Wikipedia calls it the biggest Sesotho station in South Africa with about 3.46 million listeners and a 62 in 100 audience share in the Free State.
  • Phone-in radio. Youth-focused Sesotho phone-in radio programmes were recorded for the Maseru code-switching study, so such shows contain Sesotho with English.
  • Music and street speech (kwaito, Tsotsitaal). Tsotsitaal draws on Sesotho and is the main language of kwaito music.

Sesotho varieties a tool has to cope with

  • Urban township Sesotho. Northern urban townships such as Soweto, Gauteng. Marked vocabulary differences from the Lesotho and Free State variety, due to borrowing.

Spelling, tone and names in Sesotho

Lesotho and South Africa write the same language differently, so a transcript may be correct in one orthography and look wrong in the other.

Where Sesotho is spoken

  • Lesotho. Official (one of two) and national language; about 1,493,000 speakers, 85 in 100 of the population, in 1993 data (old).
  • South Africa. One of twelve official languages; the 2011 census recorded almost four million first-language speakers, about 8 in 100 of the population, mostly in Free State and Gauteng.
  • Zimbabwe. Listed as one of sixteen official languages of Zimbabwe; no speaker figure in the source.

Transcription apps compared for Sesotho

OptionWhat its page shows about SesothoSource
AutransAutrans offers Sesotho. Measured result: Word error rate 69.4% on read speech, NCHLT test set, measured Sep 2026.autrans.online
SonixThe list on the Sonix page does not include Sesotho.sonix.ai
DescriptThe list on the Descript page does not include Sesotho.help.descript.com
TrintThe list on the Trint page does not include Sesotho.info.trint.com
MaestraThe list on the Maestra page does not include Sesotho.maestra.ai
VEEDThe list on the VEED page does not include Sesotho. It says: "Twi, Sesotho and Oromo are translation-only, so recorded as no for transcription."support.veed.io
TurboScribeThe pages read for TurboScribe say nothing about Sesotho.turboscribe.ai
RevThe pages read for Rev say nothing about Sesotho.rev.com
Happy ScribeThe pages read for Happy Scribe say nothing about Sesotho.happyscribe.com

Named: its page names the language. Not listed: its own list leaves the language out. Not stated: the pages read say nothing either way, so test a file. All pages read on 22 October 2026.

Speech APIs for developers: Sesotho

APIWhat its page shows about SesothoSource
SpeechmaticsThe list on the Speechmatics page does not include Sesotho.docs.speechmatics.com
IntronIntron names Sesotho on its page.docs.voice.intron.io
Lelapa AILelapa AI names Sesotho on its page.docs.lelapa.ai
AssemblyAIThe pages read for AssemblyAI say nothing about Sesotho.assemblyai.com
SpitchThe pages read for Spitch say nothing about Sesotho.docs.spitch.app

These are APIs for developers, not apps for a recording.

Where another tool is the better choice

  • Sonix, Descript, Trint and Maestra did not list Sesotho on the pages we read, so a file in Sesotho is a risk there until you test it.
  • For a record that must stand up in court, use a human transcriber who speaks Sesotho.

What Autrans offers for Sesotho

Word error rate 69.4% (95% confidence interval 65.7–73.6%) on read speech, NCHLT test set, measured Sep 2026. Measured on 250 recordings. Word error rate is the share of words that differ from a human-written transcript, so lower is better. Read speech means people reading set sentences aloud, so conversation, background noise and phone-quality audio will usually score worse. Test set: NCHLT speech corpus (SADiLaR).

Autrans turns audio, video, WhatsApp voice notes and phone call recordings into a transcript you can read, search, edit and export.

Languages here: Sesotho.

Questions people ask

Does Autrans transcribe Sesotho?

Yes, Sesotho is one of the live languages. Word error rate 69.4% (95% confidence interval 65.7–73.6%) on read speech, NCHLT test set, measured Sep 2026.

Do other tools name Sesotho?

Sonix, Descript, Trint and Maestra did not list Sesotho on the pages we read on 22 October 2026. A page that does not list a language may still handle it, so test a file.

Related

Start transcribing free

Send Autrans into your next meeting, use it in the room, or upload a recording. Your first 30 minutes are free, given once, and no credit card is needed to start.

Last checked 22 October 2026. How we measure accuracy. Spotted a mistake? Write to hello@autrans.online.