Skip to main content
PATCH
Correct an audio clip's transcript and word timings

Authorizations

X-CEKURA-API-KEY
string
header
required

API Key Authentication. It should be included in the header of each request.

Path Parameters

id
integer
required

A unique integer value identifying this scenario.

audio_id
string
required

Body

Validates a hand-corrected transcript for one clip.

transcript
string

What the clip actually says, as plain text. The testing agent records these words as what it said when the clip plays, and speaks them if the audio cannot be loaded. Only a clip whose status is ready can be edited; replacing its audio transcribes it again and discards the edit.

Maximum string length: 16384
words
object[]

Optional timings for the transcript, one entry per word in playback order, ending within the clip. Together the words must spell out the transcript, ignoring case, spacing and punctuation. Omit them to keep the clip's current timings when the word count is unchanged, or to spread the new words over the span the old ones covered.

Response

Read serializer for one Scenario.condition_audio entry (a plain dict).

The caller passes each entry dict with audio_id injected; fields are tolerant of missing keys. previous_action is intentionally not exposed (internal undo state for a clip that replaced a whole action).

audio_id
string | null
action
string | null
condition_id
string
read-only

First condition referencing this clip, or null when no action references it. Kept for compatibility - prefer condition_ids.

condition_ids
integer[]

Every condition referencing this clip, in authored order. Empty when the clip is uploaded but not referenced by any action yet.

steps
string
read-only

Every step referencing this clip, as paths into the conditional actions (e.g. first_message, conditions[1].then[0]).

tag
string
read-only

The reference to paste into an action, e.g.

status
string | null
transcript
string | null
words
object[]

The transcript's words with the time each is spoken in the clip. Absent until the clip has been transcribed.

duration
number<double> | null
filename
string | null
timing_source
string | null

Where the word timings came from: deepgram (measured), interpolated (spread evenly), or manual (sent with a transcript edit).

error
string | null
updated_at
string | null
playback_url
string
read-only

S3 URL of the recorded audio clip