Chinese transcription begins with two editorial decisions the broad zh tag cannot settle: which spoken variety dominates and whether readers need Simplified or Traditional characters. Record those choices, then review homophones, word segmentation, names, numbers, regional expressions, and Chinese–English switches against the audio.
Where this workflow fits
- Chinese interviews
- Lectures and seminars
- Multilingual research recordings
What to expect from Mandarin Chinese audio
The broad zh tag identifies Chinese speech but does not settle regional variety or final script convention; those choices should follow the recording and intended readers. The practical review targets are character choices, word segmentation, names, numbers, regional terms, and Simplified or Traditional style. The zh base hint does not by itself choose a regional spoken variety or Simplified versus Traditional final orthography.
Review the details that carry meaning
Names, homophones, numbers, segmentation, and Chinese–English code-switching require an explicit editorial pass. Use word-level playback for character choices, word segmentation, names, numbers, regional terms, and Simplified or Traditional style; a correction is accepted only when the surrounding clause and project convention support it.
Frozen deployment boundary
This page is published only because its language is present in capability snapshot m10-2026-07-20.1. It does not extend that catalog or claim a measured accuracy score.
- The zh base hint does not by itself choose a regional spoken variety or Simplified versus Traditional final orthography.
- Names, homophones, numbers, segmentation, and Chinese–English code-switching require an explicit editorial pass.
Reviewed provider capability
deepgram · nova-3
root zh · tiers fast, standard, precision
- Automatic detection
- Reviewed yes
- Diarization
- Reviewed yes
- Word timestamps
- Reviewed yes
Reviewed provider capability
assemblyai · universal-3-5-pro
root zh · tiers precision
- Automatic detection
- Reviewed yes
- Diarization
- Reviewed yes
- Word timestamps
- Reviewed yes
Reviewed provider capability
assemblyai · universal-2
root zh · tiers fast, standard, precision
- Automatic detection
- Reviewed yes
- Diarization
- Reviewed yes
- Word timestamps
- Reviewed yes
Questions from the workflow
How should I review a Mandarin Chinese transcript?
The broad zh tag identifies Chinese speech but does not settle regional variety or final script convention; those choices should follow the recording and intended readers. The zh base hint does not by itself choose a regional spoken variety or Simplified versus Traditional final orthography. The final pass targets character choices, word segmentation, names, numbers, regional terms, and Simplified or Traditional style, with unresolved words kept beside their source timestamps.
Reviewed sources
- IANA Language Subtag Registry · checked 2026-07-20
- AssemblyAI model and supported-language documentation · checked 2026-07-20
- AssemblyAI pre-recorded audio supported languages · checked 2026-07-20
- Unicode CLDR likely-subtags language and script data · checked 2026-07-20
- Mandarin Chinese language and orthography editorial reference · checked 2026-07-20
- Deepgram models and languages overview · checked 2026-07-20