Lao results: what to expect
Lao has very little training data in the speech model. The draft will contain many errors, and the model sometimes drifts toward Thai, which it knows far better. It works best on slow, clear speech from one speaker close to the microphone. Mor lam singing, khaen music, crowded rooms and phone calls give much weaker text.
On a long interview, a timed draft that you correct line by line is still usually quicker than typing Lao from scratch.
Upload the recording as it is: MP3, M4A, WAV, OGG, OPUS, FLAC, a voice message or a phone recording. You can also record from your microphone in the upload box. Nothing needs converting first.
Lao script and the pull toward Thai
The transcript is written in Lao script. Because Lao and Thai are close and the model has much more Thai, check for these problems:
- Whole lines written in Thai script instead of Lao. If most of the transcript comes out in Thai, re-check that Lao was selected.
- Thai spellings or Thai vocabulary in place of the Lao word the speaker used.
- Lao does not put spaces between words; spaces usually mark the end of a phrase, so long unbroken lines are normal.
- Tone marks and vowel signs are the characters most likely to be wrong.
Isan, regional accents and other languages
Vientiane Lao is closest to what the model expects. Luang Prabang and southern accents need more correction. Isan, spoken in northeast Thailand, is very close to Lao, but people there usually write in Thai script; for an Isan recording you want in Thai script, try Thai, and expect corrections either way.
Hmong, Khmu and other minority languages of Laos are not in the language list. French and English words appear in some conversations and may be written in Lao script as they sound.
The transcript opens next to a player that highlights the current line. Click a line to hear it again, correct the word in place, and your changes save automatically.
Who transcribes Lao
Typical files are voice messages from family in Laos, oral histories with Lao communities in the US, France and Australia, Buddhist talks at temples, NGO and development-project interviews, and local radio. Many users have a fluent speaker correct the draft before translating it.
Export TXT, DOCX, PDF, CSV, SRT or VTT, with timestamps and speaker names switched on or off per export. Speaker labels are available on paid plans or with a Minute Pack.
For video files, see the Lao video to text page.