Languages supported
Afrikaans benchmark accuracy (7.4% WER)
Fewer errors vs OpenAI Whisper
Afrikaans support from Otter.ai and Fathom
Van vergadering tot notule
From the call to a first draft of the minutes (notule): who said what, what was decided, and who does what next. You check it and send it.
Invite the Transkripsie bot to your Microsoft Teams, Zoom or Google Meet call, or upload a recording you already have.
Every speaker is labelled, whether they speak Afrikaans, English or both in one sentence. Our speech engine scored 92.6% accuracy on Afrikaans in a March 2025 benchmark.
The AI summary sets out what was decided and the action items, with a name next to each task where the meeting gave one. Save a summary template with your own headings, such as Besluite and Aksiepunte, and set its language to Afrikaans.
Six powerful capabilities that transform raw meetings into actionable knowledge.
Send the Transkripsie bot to a Google Meet, Zoom or Microsoft Teams call, or upload an audio or video file, and get a full transcript in 60+ languages.
Automatically detects and labels each speaker in a recording, and you can rename speakers in a click.
Get concise summaries, action items, and key decisions generated automatically after every meeting.
Search across all meetings by keyword, speaker, or topic. Find that one moment in seconds, not hours.
Transcribes 60+ languages out of the box. Afrikaans is our focus -- our speech engine scored 92.6% accuracy on Afrikaans in a March 2025 benchmark and handles Afrikaans/English code-switching.
Track talk-time ratios, meeting frequency, and engagement patterns across your entire organisation.
Point ChatGPT, Grokbot, Claude, Gemini, or Muse at your Transkripsie account. They can list your recordings and read the transcripts.
The connection is read-only. Your API key stays in the bot's secret settings, not in the chat.
How to connectTranskripsie transcribes meetings in any major language. But Afrikaans is our focus -- including Afrikaans/English code-switching, the way South Africans actually speak.
Same speech engine, same features. Supports all major world languages including:
Built on a commercial speech recognition engine benchmarked across 60 languages. Afrikaans accuracy is our specific focus.
Transkripsie transcribes 60+ languages using the same speech engine for every language. The difference: we focus on Afrikaans accuracy, code-switching, and the nuances that general-purpose tools often get wrong.
| Service | Afrikaans Accuracy | Word Error Rate | Code-Switching (AF/EN) | Auto Language Detection | Notes |
|---|---|---|---|---|---|
| Transkripsie | 92.6% | 7.4% | Focused on Afrikaans | ||
| Otter.ai | None | N/A | Only supports English, Japanese, Spanish, French | ||
| Fathom | None | N/A | 38 languages supported, Afrikaans not included | ||
| Fireflies.ai | Unverified | Unknown | Claims Afrikaans but no published accuracy, no auto language detection | ||
| OpenAI Whisper | 74.5% | 25.5% | About 3.4x the word error rate of our speech engine on Afrikaans (March 2025 benchmark) |
Accuracy figures are for the speech recognition engine Transkripsie uses, from a March 2025 speech-to-text benchmark on real-world YouTube audio across 60 languages. Results on your recordings will vary with audio quality, accents and subject matter. Competitor language support was checked against official help documentation in April 2026 and may have changed since.
The features you expect from a meeting transcription tool, in 60+ languages -- plus a particular focus on Afrikaans.
| Feature | Transkripsie | Otter.ai | Fathom | Fireflies.ai | Whisper |
|---|---|---|---|---|---|
| Transcription | |||||
| Live real-time transcription | |||||
| 60+ languages supported | |||||
| Afrikaans transcription | |||||
| 92.6% Afrikaans accuracy (7.4% WER) | |||||
| Afrikaans/English code-switching | |||||
| Auto language detection | |||||
| File upload transcription | |||||
| Meeting Intelligence | |||||
| Auto-join Google Meet | |||||
| Meeting video recording | |||||
| Video playback with transcript | |||||
| Speaker labels | |||||
| Speaker renaming | |||||
| AI-generated summaries | |||||
| Action item extraction | |||||
| Recording disclosure in meeting | |||||
| Search & Analytics | |||||
| Cross-meeting smart search | |||||
| Meeting analytics dashboard | |||||
| Talk-time ratios per speaker | |||||
| Daily activity tracking | |||||
| Platform & Access | |||||
| Web dashboard | |||||
| Chrome extension | |||||
| API access | |||||
| Audio playback with transcript | |||||
| Segment-level audio clips | |||||
| Self-hosted / on-premise option | |||||
| Multilingual team support | |||||
| Total features | 23/26 | 15/26 | 12/26 | 17/26 | 6/26 |
Feature score
Transkripsie features in this comparison
As of April 2026, Otter.ai supported four languages: English (US and UK), Japanese, Spanish, and French. There was no Afrikaans support and no announced plan to add it, so Afrikaans speech in your meetings would not be transcribed.
As of April 2026, Fathom supported transcripts in 38 languages, including Dutch and German, but Afrikaans was not one of them. Despite being closely related to Dutch, Afrikaans needs its own language model, and Fathom had not announced Afrikaans support.
As of April 2026, Fireflies listed Afrikaans among its 100+ supported languages, but published no accuracy figures for Afrikaans specifically. It also lacked automatic language detection, meaning users had to select the language for each meeting -- a significant limitation for bilingual Afrikaans/English conversations.
OpenAI Whisper (used by many transcription tools under the hood) had a 25.5% Word Error Rate on Afrikaans in the March 2025 benchmark -- roughly one in four words wrong. That is 3.4x more errors than the 7.4% WER of the speech engine Transkripsie uses. For a 30-minute meeting, that can mean hundreds of additional errors in the transcript.
Leave your email and we'll let you know when you can start. Speaker-labelled transcriptions in 60+ languages, with a particular focus on Afrikaans.