Voice output, not transcription
This workflow creates spoken replies from text. If your agent also needs to hear users, use a separate speech-to-text layer for input, then pass approved response text into TextToSpeechSkills for output.
For agent builders
Let agents create speech without broad account access
TextToSpeechSkills provides a narrow, reviewable speech workflow through the API today. Scoped MCP access for validating markup, choosing templates, checking credits, and creating jobs follows after npm publication.
Who is this for?
TextToSpeechSkills gives product backends a safe API for voice output without broad account access. After npm publication, the MCP server will expose focused tools for validating natural expression markup, selecting approved voice templates, checking credit use, creating speech jobs, and returning audio URLs. This is speech output, not a speech-to-text or transcription API.
Easy LLM setup
LLM-ready even for non-technical teams
Use the studio for first tests or add a scoped key to an API integration. MCP installation and approved-template access will be enabled after npm publication.
Read setup guide01Create a scoped key
02Use the studio or API
03Choose a voice template
04Generate and review audio
Tool calls you can review
Agents use clear speech tools instead of improvising HTTP calls or handling hidden settings.
Templates protect quality
Approved voice templates keep agent output consistent across users and workflows.
Usage stays visible
Credit previews, job states, and workspace billing make automated audio easier to manage.
When this helps
Agent builders, automation teams, and AI product teams usually need a repeatable path for writing, review, generation, billing, and reuse. The most important jobs here are voice output, not transcription, tool calls you can review, templates protect quality, usage stays visible. Those are the moments where voice becomes part of real work instead of a one-off export.
How the workflow works
Start with readable text, add natural-language expression directions when tone matters, choose an approved voice template, and create a speech job through the UI or API. After npm publication, MCP will use the same pattern for AI agent voice output, voice agent text-to-speech, speech output for voice agents, helping humans and LLM apps share one process without exposing internal routing or credentials.
Before you roll it out
Decide which templates are approved, how natural expression markup should be reviewed, who can create workspace keys, and which usage limits are acceptable. Those choices keep automated voice generation useful without letting it sprawl from the first paid Test plan through Pro, Scale, and Business usage.
Common questions
What teams usually ask before starting
These are the practical details that matter before a team adds speech generation to a real workflow.
Who should use Voice Output for AI Agents and Voice Apps?
Agent builders, automation teams, and AI product teams should use this page when they want generated speech that is easy to review, consistent across prompts, and simple to connect to LLM tools. The core workflow combines natural expression markup, voice templates, credit previews, and job-based generation.
Can a non-technical user connect this to an LLM app?
Use the studio for first tests or add a scoped key to an API integration. MCP installation and approved-template access will be enabled after npm publication. The setup guide keeps the first path short while still giving developers a clean API when the workflow moves into a product backend.
How does pricing stay predictable?
Every paid plan uses credits. Teams can add credit packs when needed, and workspaces on Pro and higher add central billing for $2 per user per month.
API playground
Plain JSON in, speech job out
{
"text": "[quiet] hello. [loud and angry] how are you?",
"voice_template": "vt_calm_narrator_v1",
"format": "wav"
}Job created200 audio ready
MCP package
MCP package release pending
npm release status
Install commands are not available yetThe package is prepared for release but is not published. Use the browser studio or API today; MCP commands will appear here after npm publication.
Read API docs