Reconcile two text inputs against audio

POST /v1/audio/diff-alignments

Unreleased. Requires diff_alignment capability. Submit both arrays as JSON strings. No single-input fallback, normalization, extended search or streaming. Result timestamps include offset; original_index refers to the submitted reference array (-1 means unattributed).

Guide and examplesForced alignment All endpoints

Request body

multipart/form-dataRequired

filefileRequiredAudio window, at most 300 seconds including padding.

reference_segmentsstringRequiredJSON array of nonempty text and optional speaker objects.

transcript_segmentsstringRequiredJSON array of nonempty text and optional speaker objects.

modelstring

languagestring

offsetnumberAbsolute timestamp offset in seconds.≥ 0Default 0

Responses

200Native acoustic diff result. Speaker labels come from input text, not verified voice identities.

durationnumber

offsetnumber

speaker_sourcestringinput_text

segmentsobject[]

textstring

startnumber

endnumber

speakerstring | null

original_indexinteger≥ -1

wordsobject[]

400Invalid or unsupported request fields.

401Authentication or authorization required.

413Audio exceeds 300 seconds or the request size limit.

422Acoustic alignment failed.

503Runtime unavailable.