AiPlannerVoiceDef | Metadata API Developer Guide

Parent Type

This type extends the Metadata metadata type and inherits its fullName field.

File Suffix and Directory Location

AiPlannerVoiceDef components have the suffix .aiPlannerVoiceDef and are stored in the aiPlannerVoiceDef folder.

Version

AiPlannerVoiceDef components are available in API version 63.0 and later.

Special Access Rules

AiPlannerVoiceDef is available only if Agents and Agentforce Voice are enabled in your org.

Fields

Field Name	Description
additionalConfigs	Field Type string Description A JSON string for extensible configuration settings not covered by explicit fields. Allows additional voice parameters to be stored without requiring schema changes.
developerName	Field Type string Description Required. Represents the API name of the voice definition. This name must be unique across all voice definitions. Can contain only underscores and alphanumeric characters and must be unique in your org. It must begin with a letter, not include spaces, not end with an underscore, and not contain two consecutive underscores.
inboundFillerWordsBool	Field Type boolean Description Specifies whether filler words (such as "um" or "uh") are included when processing inbound audio. When true, filler words are preserved in the transcription output. When false, they are stripped.
inboundKeywords	Field Type string Description A JSON object of keywords or phrases with optional boost weights that improve speech recognition accuracy for inbound audio. Positive values increase recognition priority; negative values suppress. Example: {"keywords": {"Salesforce": 5, "Agentforce": 5, "cancel": -10}}.
inboundModel	Field Type string Description The speech-to-text model used to transcribe inbound audio. References a deployed model record in MktMLAIModel (e.g., sfdc_ai__Default_Nova_2_PhoneCall).
label	Field Type string Description The user-facing label for this AiPlannerVoiceDef component as it appears in the Salesforce UI.
languageCode	Field Type string Description The BCP 47 language code specifying the language used for voice processing, for example en-US for American English or ja-JP for Japanese.
outboundFillerSentences	Field Type string Description A JSON object of filler sentences the voice agent speaks while processing a response, organized by category. Helps maintain natural conversational flow during pauses. Example: {"fillerSentences": {"waiting": ["Let me look into that...", "Just a moment..."], "confirmation": ["Got it, let me confirm that."], "fallback": ["Sorry, could you repeat that?"]}}.
outboundModel	Field Type string Description The text-to-speech model used to synthesize outbound audio. References a deployed model record in MktMLAIModel (e.g., sfdc_ai__Default_Eleven_Flash_V2).
outboundSimilarity	Field Type double Description A value between 0 and 1 that controls how closely the synthesized voice matches the selected voice profile. Higher values produce a more faithful reproduction; lower values allow more natural variation. Default is typically 0.75.
outboundSpeedNum	Field Type double Description A multiplier that controls the speaking rate of the outbound voice. A value of 1.0 represents normal speed. Typical range is 0.7 (slower) to 1.2 (faster), depending on the voice persona's configured bounds.
outboundStability	Field Type double Description A value between 0 and 1 that controls the consistency of the outbound voice tone and delivery. Higher values produce a more stable, predictable output; lower values introduce more expressive variation. Default is typically 0.65.
outboundStyleExaggeration	Field Type double Description A value between 0 and 1 that amplifies the speaking style of the selected voice. Higher values increase expressiveness and emotional range in the synthesized speech. A value of 0 produces neutral delivery.
outboundVoice	Field Type string Description The name of the voice persona used for text-to-speech synthesis in outbound audio (e.g., elevenlabs-Mark-en_US). Maps to a voice configuration defined in the voice metadata layer.
pronunciationDict	Field Type string Description A custom pronunciation dictionary that defines phonetic overrides for how specific words or phrases are pronounced by the text-to-speech engine. Useful for brand names, technical terms, or unusual proper nouns.
voiceIdentifier	Field Type string Description A unique identifier for the voice configuration, typically the pre-computed hash ID from the voice metadata layer (e.g., f64af13e6fb0). Used to reference or distinguish this voice setup across planner configurations.

Declarative Metadata Sample Definition

The following is an example of an AiPlannerVoiceDef component.

1<?xml version="1.0" encoding="UTF-8"?>
2<GenAiPlannerBundle xmlns="http://soap.sforce.com/2006/04/metadata">
3    <description>This agent is either testing you, or you&apos;re testing it.</description>
4    <masterLabel>My Test Agent</masterLabel>
5    <plannerType>AiCopilot__ReAct</plannerType>
6    <voiceDefinition>
7        <developerName>dev2Automation</developerName>
8        <inboundFillerWordsBool>false</inboundFillerWordsBool>
9        <inboundKeywords>key2</inboundKeywords>
10        <inboundModel>sfdc_ai__DefaultOpenAIGPT35Turbo</inboundModel>
11        <label>label2</label>
12        <outboundFillerSentences>test2</outboundFillerSentences>
13        <outboundModel>sfdc_ai__DefaultOpenAIGPT35Turbo</outboundModel>
14        <outboundSimilarity>0.8</outboundSimilarity>
15        <outboundSpeedNum>1.2</outboundSpeedNum>
16        <outboundStability>0.5</outboundStability>
17        <outboundStyleExaggeration>1.8</outboundStyleExaggeration>
18        <outboundVoice>test2</outboundVoice>
19        
20    </voiceDefinition>
21</GenAiPlannerBundle>

The following is an example package.xml that references the previous definition.

1<?xml version="1.0" encoding="UTF-8"?>
2<Package xmlns="http://soap.sforce.com/2006/04/metadata">
3    <types>
4        <members>*</members>
5        <name>GenAiPlannerBundle</name>
6    </types>
7    <version>63.0</version>
8</Package>