Skip to content
KortexyaKortexyaReasoning Layer SDK

Interface: EnrollSpeakerRequest

@kortexya/reasoninglayer

@kortexya/reasoninglayer / EnrollSpeakerRequest

Interface: EnrollSpeakerRequest

Section titled “Interface: EnrollSpeakerRequest”

Defined in: src/types/speakers.ts:12

Body of POST /api/v1/speakers/enroll — enroll (or re-enroll) a speaker.

Properties

Section titled “Properties”

audio

Section titled “audio”

audio: string

Defined in: src/types/speakers.ts:14

Base64-encoded audio sample of the speaker (≥ ~3 s of speech).

audioMime?

Section titled “audioMime?”

optional audioMime?: string | null

Defined in: src/types/speakers.ts:16

MIME hint (e.g. "audio/wav") for decoding the sample. Optional.

name

Section titled “name”

name: string

Defined in: src/types/speakers.ts:21

Display name for this speaker. Re-enrolling the same (name, variant) pair (case-insensitive) overwrites the existing profile.

variant?

Section titled “variant?”

optional variant?: string | null

Defined in: src/types/speakers.ts:31

Optional sample label, so one person can hold SEVERAL embeddings ("15-s read-aloud", "meeting 07-30", "sung sample").

Remarks

Section titled “Remarks”

A voice is not one point in embedding space: a sung sample is what makes that artist’s other songs match. Several prints that share a name resolve each cluster to the best cosine overall.