Add support for a custom word pronunciation dictionary that allows users to define how specific words (such as brand names) should be pronounced by the AI voice agent, similar to the functionality offered by HeyGen and ElevenLabs.

Problem Statement:

Currently, voice agents rely on default Text-to-Speech (TTS) pronunciation rules. This causes incorrect pronunciation of brand names, proprietary terms, and industry-specific words.

For example, our brand name Soccerfy must be pronounced “Soc-ker–fye”, but TTS engines often pronounce it incorrectly (e.g., “Soc-ker-fee”). This leads to:

There is currently no reliable way to enforce consistent pronunciation across all voice interactions.

Proposed Solution:

Introduce a Custom Pronunciation Dictionary that allows users to:

- Phonetic spelling (e.g., “Soc-ker-fye”)

- IPA or SSML phoneme notation (optional)

- Voice agents

- Call scripts

- Dynamic AI responses

This dictionary would override the default TTS behavior whenever the custom word appears.

Example Use Case:

Custom Word: Soccerfy

Display Text: Soccerfy

Pronunciation: soc-ker-fye

Whenever the AI agent speaks “Soccerfy,” it should consistently pronounce it as “Soc-ker-fye” without requiring manual script changes or workarounds.

Comparable Implementations:

HeyGen: Custom pronunciation entries for brand and proper nouns

ElevenLabs: Pronunciation dictionaries with phoneme-level control

These platforms demonstrate that this feature is technically feasible and significantly improves voice consistency.

Business Impact:

Suggested Priority:

High — essential for production-grade voice agents and branded AI deployments.

Optional Enhancements (Nice to Have):