Settings & Pronunciation

These endpoints let you read and update the server-persisted defaults that drive the PeachDesk Voice Studio experience, and manage the pronunciation dictionary that the Voice Engine applies before synthesis. All routes are exposed by the PeachDesk API under the /api/voice base path and authenticated with your PeachDesk API key in the X-API-Key header.

Settings are persisted per workspace on PeachDesk’s servers. They act as the defaults the Voice Studio UI loads on open — individual generation or capture requests can still override most of these values inline.

Authentication

Every request must include your PeachDesk API key in the X-API-Key header (an interactive session token in Authorization: Bearer also works — see Authentication):

X-API-Key: bibha_your_api_key

Capture settings

Defaults for the Captures flow — speech-to-text transcription and optional LLM refinement. Models referenced here (STT and refinement LLM) run on PeachDesk’s GPU infrastructure; you select them by identifier and never download or manage any model files locally.

Get capture settings

GET /api/voice/settings/captures Returns the current capture and refinement defaults for the workspace.

string

Speech-to-text model identifier — lipi (the default), lipi-bharat or lipi-live. The legacy size keys base, small, medium, large, and any other pre-consolidation identifier are still accepted so older settings rows keep loading, but they resolve to LIPI.

string

Source language code, or auto to let the engine detect it. Default auto.

string

transcribe (same language) or translate (to English). Default transcribe.

integer

Beam search width, 1–5. Higher is more accurate but slower. Default 1.

string

Optional biasing vocabulary (up to 2000 chars) of domain terms to favour. Default "".

boolean

Automatically run LLM refinement on a new capture’s transcript. Default true.

string

Refinement model. BRAHMA-30B-FP8 (BRAHMA-30B) is the only accepted value — the refinement model is pinned server-side.

boolean

Remove filler words and tidy punctuation during refinement. Default true.

boolean

Resolve spoken self-corrections (“no wait, I mean…”). Default true.

boolean

Preserve code, identifiers, and technical terms verbatim. Default true.

boolean

Allow the refined transcript to flow into the active composer automatically. Default true.

string | null

Voice profile used to read captures back aloud, if set.

boolean

Enable the in-app capture shortcut while the Voice Studio is open. Default false.

string[]

Key chord (1–6 keys) used as push-to-talk inside the Voice Studio.

string[]

Key chord (1–6 keys) used to toggle dictation on and off inside the Voice Studio.

curl https://app.peachdeskai.com/api/voice/settings/captures \
  -H "X-API-Key: $PEACHDESK_API_KEY"

{
  "stt_model": "lipi",
  "language": "auto",
  "stt_task": "transcribe",
  "stt_beam_size": 1,
  "stt_vocabulary": "",
  "auto_refine": true,
  "llm_model": "BRAHMA-30B-FP8",
  "smart_cleanup": true,
  "self_correction": true,
  "preserve_technical": true,
  "allow_auto_paste": true,
  "default_playback_voice_id": null,
  "hotkey_enabled": false,
  "chord_push_to_talk_keys": [],
  "chord_toggle_to_talk_keys": []
}

Update capture settings

PUT /api/voice/settings/captures Partial update — send only the fields you want to change. Every field is optional; omitted fields keep their current value. Returns the full, updated capture settings object.

string

string

Source language code, or auto.

string

transcribe | translate.

integer

1–5.

string

Biasing vocabulary, up to 2000 chars.

boolean

string

BRAHMA-30B-FP8 — the only accepted value.

boolean

string

boolean

string[]

1–6 keys.

string[]

1–6 keys.

curl -X PUT https://app.peachdeskai.com/api/voice/settings/captures \
  -H "X-API-Key: $PEACHDESK_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "stt_model": "lipi",
    "language": "en",
    "auto_refine": true,
    "llm_model": "BRAHMA-30B-FP8"
  }'

Generation settings

Defaults for the Generation tab — how long text is chunked and how output audio is finished.

Get generation settings

GET /api/voice/settings/generation Returns the current text-to-speech generation defaults.

integer

Max characters per chunk when splitting long text. 100–5000. Default 800.

integer

Crossfade between chunks, in milliseconds (0 for a hard cut). 0–500. Default 50.

boolean

Normalize output loudness. Default true.

boolean

Automatically play the take when generation completes. Default true.

curl https://app.peachdeskai.com/api/voice/settings/generation \
  -H "X-API-Key: $PEACHDESK_API_KEY"

{
  "max_chunk_chars": 800,
  "crossfade_ms": 50,
  "normalize_audio": true,
  "autoplay_on_generate": true
}

Update generation settings

PUT /api/voice/settings/generation Partial update — send only the fields you want to change. Returns the full, updated generation settings object.

integer

100–5000.

integer

0–500.

boolean

curl -X PUT https://app.peachdeskai.com/api/voice/settings/generation \
  -H "X-API-Key: $PEACHDESK_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{ "max_chunk_chars": 1200, "crossfade_ms": 80 }'

Voice Activity Detection (VAD) settings

Tuning for the speech-detection that trims silence around recorded audio in the Captures and Talk flows.

Get VAD settings

GET /api/voice/settings/vad Returns the current VAD preferences.

boolean

Whether VAD trimming is active. Default false.

number

Speech-probability threshold, 0.0–1.0. Higher is stricter. Default 0.5.

integer

Minimum speech duration to keep a segment, in ms. 0–5000. Default 250.

integer

Minimum silence to split segments, in ms. 0–5000. Default 100.

integer

Padding added around detected speech, in ms. 0–1000. Default 30.

integer

Audio sample rate the VAD operates at, in Hz. Default 16000.

curl https://app.peachdeskai.com/api/voice/settings/vad \
  -H "X-API-Key: $PEACHDESK_API_KEY"

{
  "enabled": false,
  "threshold": 0.5,
  "min_speech_ms": 250,
  "min_silence_ms": 100,
  "speech_pad_ms": 30,
  "sample_rate": 16000
}

Update VAD settings

PUT /api/voice/settings/vad Partial update — send only the fields you want to change. Returns the full, updated VAD settings object.

boolean

number

0.0–1.0.

integer

0–5000.

integer

0–5000.

integer

0–1000.

integer

Sample rate in Hz.

curl -X PUT https://app.peachdeskai.com/api/voice/settings/vad \
  -H "X-API-Key: $PEACHDESK_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{ "enabled": true, "threshold": 0.6, "speech_pad_ms": 50 }'

Pronunciation dictionary

Pronunciation rules tell the Voice Engine how to say specific terms before synthesis. A rule is either an alias (rewrite the term to a spelled-out replacement) or a phoneme override. Rules are either global (apply to every voice) or profile-scoped (apply only to one voice profile).

Use an alias rule for acronyms or brand names (“PeachDesk” → “Peach Desk”), and a phoneme rule when you need precise control over the sound of a term.

List pronunciation rules

GET /api/voice/pronunciation Lists pronunciation rules, optionally filtered by scope and/or profile.

string

Filter by scope: global or profile.

string

Filter by the voice profile a rule belongs to.

Returns an array of rule objects:

string

Unique rule identifier.

string

global or profile.

string | null

Voice profile the rule applies to, for profile-scoped rules.

string

The term the rule matches.

string

alias or phoneme.

string | null

Spelled-out replacement, for alias rules.

string | null

Phoneme string, for phoneme rules.

boolean

Whether matching is case-sensitive. Default false.

boolean

Whether the rule is active. Default true.

string

ISO-8601 creation timestamp.

curl "https://app.peachdeskai.com/api/voice/pronunciation?scope=global" \
  -H "X-API-Key: $PEACHDESK_API_KEY"

[
  {
    "id": "rule_a1b2c3",
    "scope": "global",
    "profile_id": null,
    "term": "PeachDesk",
    "rule_type": "alias",
    "replacement": "Peach Desk",
    "phoneme": null,
    "case_sensitive": false,
    "enabled": true,
    "created_at": "2026-06-23T10:15:00Z"
  }
]

Create a pronunciation rule

POST /api/voice/pronunciation Creates a global or profile-scoped rule. Returns the created rule object.

string

required

The term to match (1–200 chars).

string

default:"alias"

alias or phoneme.

string

Spelled-out replacement for alias rules (up to 400 chars).

string

Phoneme override for phoneme rules (up to 400 chars).

string

default:"global"

global or profile.

string

Required when scope is profile — the voice profile the rule applies to.

boolean

default:"false"

boolean

default:"true"

A 400 error is returned if the rule is invalid — for example, a profile-scoped rule without a profile_id, or a rule missing the value appropriate to its rule_type.

curl -X POST https://app.peachdeskai.com/api/voice/pronunciation \
  -H "X-API-Key: $PEACHDESK_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "term": "GIF",
    "rule_type": "alias",
    "replacement": "jiff",
    "scope": "global"
  }'

{
  "id": "rule_d4e5f6",
  "scope": "global",
  "profile_id": null,
  "term": "GIF",
  "rule_type": "alias",
  "replacement": "jiff",
  "phoneme": null,
  "case_sensitive": false,
  "enabled": true,
  "created_at": "2026-06-23T10:20:00Z"
}

Update a pronunciation rule

PUT /api/voice/pronunciation/ Partial update of an existing rule. Returns the updated rule object, or 404 if the rule does not exist.

string

required

The rule to update.

string

1–200 chars.

string

alias or phoneme.

string

Up to 400 chars.

string

Up to 400 chars.

boolean

curl -X PUT https://app.peachdeskai.com/api/voice/pronunciation/rule_d4e5f6 \
  -H "X-API-Key: $PEACHDESK_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{ "enabled": false }'

Delete a pronunciation rule

DELETE /api/voice/pronunciation/ Deletes a rule. Returns a confirmation message, or 404 if the rule does not exist.

string

required

The rule to delete.

curl -X DELETE https://app.peachdeskai.com/api/voice/pronunciation/rule_d4e5f6 \
  -H "X-API-Key: $PEACHDESK_API_KEY"

{ "message": "Pronunciation rule deleted" }

Resources

Authentication & Errors

Voice Engine API

Bibha BRAHMA LLM

Bibha VAANI TTS

Bibha LIPI STT

Settings & Pronunciation

Authentication

Capture settings

Get capture settings

Update capture settings

Generation settings

Get generation settings

Update generation settings

Voice Activity Detection (VAD) settings

Get VAD settings

Update VAD settings

Pronunciation dictionary

List pronunciation rules

Create a pronunciation rule

Update a pronunciation rule

Delete a pronunciation rule

​Authentication

​Capture settings

​Get capture settings

​Update capture settings

​Generation settings

​Get generation settings

​Update generation settings

​Voice Activity Detection (VAD) settings

​Get VAD settings

​Update VAD settings

​Pronunciation dictionary

​List pronunciation rules

​Create a pronunciation rule

​Update a pronunciation rule

​Delete a pronunciation rule

Authentication

Capture settings

Get capture settings

Update capture settings

Generation settings

Get generation settings

Update generation settings

Voice Activity Detection (VAD) settings

Get VAD settings

Update VAD settings

Pronunciation dictionary

List pronunciation rules

Create a pronunciation rule

Update a pronunciation rule

Delete a pronunciation rule