> ## Documentation Index
> Fetch the complete documentation index at: https://docs.unpod.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Voice Profiles

> Browse and select the speech stack for your agents.

## What Is a Voice Profile?

A [Voice Profile](https://docs.unpod.ai/get-started/core-concepts.md#voice-profile) bundles three
things into a single reusable configuration:

* **Speech-to-Text (STT)** - transcribes the caller's audio into text
* **Text-to-Speech (TTS)** - converts agent text into natural-sounding audio
* **Voice** - the specific voice identity (gender, accent, character) used for TTS output

When you attach a voice profile to an agent, every call that voice handles uses that configuration automatically - including failover if a provider goes down.

Voice profile pipeline

Caller speaksAudio in

STTFrom the voice profileuser textDialog machineYour agent logicagent textTTS + voiceFrom the voice profile

Caller hearsAudio out

Unpod runs itYou run itCaller

Voice profiles are managed by Unpod. You browse and select from the available list - you do not configure STT/TTS providers directly.

***

## Listing Voice Profiles via SDK

```python theme={null}
import asyncio
from unpod import AsyncClient

async def main():
    async with AsyncClient() as client:  # reads credentials from .env
        # All profiles
        profiles = await client.voice_profiles.list()

        # Filter by language
        en_profiles = await client.voice_profiles.list(language="en")
        es_profiles = await client.voice_profiles.list(language="es")

        for p in profiles:
            print(
                p.profile_id,
                p.name,
                f"gender={p.gender}",
                f"quality={p.quality}",
            )

asyncio.run(main())
```

### VoiceProfile fields

| Field | Type | Description |
| - | - | - |
| `profile_id` | `str` | Profile ID to pass as `voice_profile` when creating an agent |
| `name` | `str` | Human-readable name, e.g. `"Emma (HD)"` |
| `gender` | `str` | `male`, `female`, or `neutral` |
| `quality` | `str` | `standard` or `high` |
| `languages` | `list[str]` | Supported language codes, e.g. `["en", "hi"]` |
| `description` | `str \| None` | Short description of the voice character |
| `latency_ms` | `int \| None` | Expected end-to-end latency in milliseconds |
| `greeting_message` | `str \| None` | Default greeting for this profile |

***

## Getting a Single Profile

```python theme={null}
import asyncio
from unpod import AsyncClient

async def main():
    async with AsyncClient() as client:  # reads credentials from .env
        profile = await client.voice_profiles.get("vp_en_female_hd")
        print(profile.name, profile.quality, profile.languages)

asyncio.run(main())
```

***

## Choosing the Right Profile

### By use case

| Use case | Recommended qualities |
| - | - |
| Customer support | High quality, natural voice, low latency |
| Sales outreach | Expressive, high quality, brand-matched |
| Appointment booking | Standard quality - lower latency |
| Multilingual | Match language codes to caller's region |

### By latency

Total call latency has three components: STT transcription + LLM thinking + TTS synthesis.

* Choose `quality == "high"` when voice naturalness matters more than speed
* Choose `quality == "standard"` for latency-sensitive or high-volume deployments
* Check `latency_ms` on the profile object for the estimated end-to-end figure

***

## Using a Profile When Creating an Agent

Pass `profile_id` as `voice_profile` when you create the agent:

```python theme={null}
import asyncio
from unpod import AsyncClient, Prompt

async def main():
    async with AsyncClient() as client:  # reads credentials from .env
        # List to find the right profile
        profiles = await client.voice_profiles.list(language="en")
        hd_female = next(
            p for p in profiles if p.quality == "high" and p.gender == "female"
        )

        voice = await client.agents.voice.create(
            "support-bot",
            brain=Prompt("You are a concise support assistant."),
            name="Support Bot",
            voice_profile=hd_female.profile_id,
        )
        print("Created agent with profile:", voice.voice_profile_name)

asyncio.run(main())
```

***

## Giving an Agent a Second Voice

One agent, one brain, many voices - a second language is a second voice, not a
second agent. The new voice inherits the brain unchanged.

```python theme={null}
import asyncio
from unpod import AsyncClient

async def main():
    async with AsyncClient() as client:  # reads credentials from .env
        await client.agents.voice.add(
            "support-bot",
            voice_profile="vp_es_male_std",     # Spanish male
            greeting="Hola, gracias por llamar.",
        )

        agent = await client.agents.get("support-bot")
        for v in agent.voices:
            print(v.voice_profile_id, v.voice_profile_name, v.is_default)

        # Drop one voice; the agent keeps the rest
        await client.agents.voice.remove("support-bot", "vp_es_male_std")

asyncio.run(main())
```

***

## Next Steps

> [card] [Agents](https://docs.unpod.ai/speech-stack/agents.md)
>
> Attach a voice profile when creating an agent, and add more voices later.

> [card] [AgentRunner & Sessions](https://docs.unpod.ai/agent-sdk/agent-runner.md)
>
> Install the SDK and configure your runner process.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.