Audio Project Manager pairs AI transcription with careful speech

Audio Project Manager 4.6.0, released on 1 October, adds something valuable to a transcription workflow: another person speaking. Its new BOLD route includes Careful Speech, an oral translation into a more widely used language, and clause-level text transcription.[1]

The release also updates automatic transcription, tying it to the language selected for that workflow step and supporting related languages. Waveform zoom and segment playback help workers move through the recording.[1]

Listen, repeat, translate

BOLD stands for Basic Oral Language Documentation. In Steven Bird’s 2010 paper, describing work with Usarufa speakers in Papua New Guinea, the method used two people and two recorders. They first listened to the complete story and discussed unclear vocabulary. One person then controlled playback while the other repeated short phrases slowly and accurately.[2]

Translation was a separate pass, using longer chunks corresponding to complete ideas. The second recorder kept running, capturing the original audio, the new spoken version and any corrections. That preserved explanations that could disappear from a clean written account.[2]

The pilot revealed why preparation mattered: younger speakers sometimes did not know words used by elders, and long segments could turn into paraphrases. Listening together beforehand and separating playback from speaking helped resolve those problems.[2]

APM makes that layered process an explicit software workflow, bringing careful speech, wider-language translation and clause transcription into the same project. The contribution is coordination: teams can work through several forms of the story together.[1]

For teams learning the technique, the Endangered Languages Documentation Programme offers free SayMore for BOLD training. It covers careful-speech recordings, oral translations and linked ELAN annotation files. This is a separate software route, with guides for understanding and practising the method.[3]

The extra voice becomes part of the record: pronunciation, meaning and the discussion needed to make both understandable.[2]

Report an error or suggest a correction