Via cyeam asr call the cyeam.com speech recognition interface (the server proxies to the ASR model) to convert local audio into text. No login or configuration required.
Usage
cyeam asr <audio-file>
cyeam --pretty asr <audio-file>
By default outputs a JSON envelope,data containing file and text；--pretty outputs only the recognized text.
Constraints
- Supports common audio formats: m4a, mp3, wav, webm, ogg, aac, flac, amr, 3gp, opus; sets the upload Content-Type according to the extension.
- A single file must not exceed 15MB (server limit); exceeding it will directly return an error.
- Recognition requires network access; when the interface is unavailable the command returns an error—do not assemble HTTP requests yourself as a substitute.
Recognition quality notes
- For children's speech and recordings with background voices or noise, the recognized text may contain noise content or typos. When grading homework, use the recognized text as reference evidence and judge in light of the original question's requirements; segments that are clearly background voices (such as a parent speaking) should not be counted as the child's reading-aloud errors.
- When there is only one recording and no original text to compare against (such as a listening comprehension question), you cannot use it to judge whether the content is correct; you can only confirm whether the reading was completed.
Homework scenarios
Trello homework recordings (homework-audio-*.m4a) in the cyeam trello homework report automatically include the transcript field (English cards only), which can generally be read directly; when the report has no recognized text or recognition failed, use this command to manually transcribe the attachment's saved_to path. See the homework skill for the detailed process.