Transcribe Welsh audio with WhisperAI
WhisperAI transcribes Welsh (Cymraeg) audio for the around 900,000 speakers who use the language across Wales, with smaller communities in Patagonia (Argentina) and the Welsh diaspora. Built on OpenAI Whisper and tuned for the realities of Welsh speech, the service produces publication-grade transcripts in Latin alphabet with circumflex (ô, â, ŵ, ŷ, ê, î, û).
Who actually needs Welsh transcription
Welsh demand comes from Welsh-government documentation (Welsh and English have equal legal status in Wales), S4C broadcasting, Welsh-medium education, and the active Welsh-revitalisation community.
Dialects and varieties handled
- Northern Welsh
- Southern Welsh
- Patagonian Welsh
Welsh mutations and the difference between northern and southern speech
Welsh consonants mutate at the start of words depending on grammatical context (treigladau) — "cath" (cat) becomes "gath" or "chath" or "nghath" depending on what comes before. ASR that doesn't understand Welsh mutations produces wrong words on every other phrase. Whisper handles the full mutation system and the regional split between northern and southern Welsh.
Northern Welsh, Southern Welsh and Patagonian Welsh are all supported.
How Welsh speakers actually use WhisperAI
- Welsh-government documentationWelsh Government and Senedd offices transcribe Welsh-language meetings under Welsh-language standards requirements.
- S4C broadcastingS4C transcribes Welsh-language broadcasts for caption use and editorial archives.
- Welsh-medium educationWelsh-medium schools and Welsh-language degree programmes transcribe instructional and administrative content.
- Welsh academic and cultural workUniversities and cultural institutions document Welsh oral histories and qualitative research.
What Welsh transcripts include
Native Latin alphabet with circumflex (ô, â, ŵ, ŷ, ê, î, û)
Output in Latin alphabet with circumflex (ô, â, ŵ, ŷ, ê, î, û) with all diacritics, tone marks, special characters and script-specific conventions preserved — never transliteration.
Speaker labels
Diarisation that separates and labels speakers in Welsh interviews, panels and multi-party meetings.
Timestamps and SRT
Word-level timestamps and SRT subtitle export for Welsh video captioning, with proper line-breaking for the script.
Editor and exports
In-browser editor with PDF, DOCX, TXT and SRT exports. Edit the transcript without leaving the browser, then download the format your downstream workflow expects.
How to transcribe Welsh audio
Upload your Welsh recording
Drop in MP3, WAV, M4A, MP4 or similar — files up to 5GB. Long-form Welsh interviews, lectures and meetings work without splitting.
Whisper transcribes the audio
The model recognises Welsh speech across the dialects above and outputs Latin alphabet with circumflex (ô, â, ŵ, ŷ, ê, î, û). Code-switched English and other languages in the same recording are handled inline.
Edit, label and export
Open the transcript in the editor, fix any names or terms, label speakers, and export to PDF, DOCX, TXT or SRT. Optional AI summary surfaces the key points and decisions.
Transcribe in 100+ languages
WhisperAI supports Welsh alongside over 100 other languages with the same accuracy and editor experience.
Other widely-used pages: English, Spanish, French, German, Portuguese, Japanese, Korean, Chinese, Arabic, Russian, Hindi, Italian, and 80+ more languages including Swedish, Norwegian, Danish, Finnish, Greek, Malay, Filipino and beyond.
See every supported language, with transcription, realtime and translation coverage.
Transcription guides and best practices
Start transcribing Welsh today
Sign up free, drop in your first Welsh file, and have a usable transcript in minutes.