aethexDocs
Dashboard

Voices

Aethex voices are language-specific TTS models. Browse the catalog, preview any voice, then assign one to your agent.

Languages

Every voice is tagged with a language. Make sure the voice language matches the agent's language field — using an English voice on a French agent gives odd accent mixing.

Voice metadata

Each voice in GET /voices carries metadata you can filter and display on: name, language, gender, and supports_dialect_style. Use these to build a picker or to narrow the catalog before previewing.

Gender

Every voice reports a gender of male, female, or neutral. There is no dedicated query parameter — filter client-side on the field returned by GET /voices.

Tags

Voices can be labelled with tags drawn from a fixed vocabulary. Filter to a single tag with ?tag=, and fetch the full set of valid tokens from GET /voices/tag-vocabulary (requires the voices:read scope) — tokens outside that set aren't accepted. No catalog voices are tagged yet, so tag filters currently return an empty list.

Dialect style

Some voices can speak a localized dialect in addition to their base language. These report supports_dialect_style: true. Pair such a voice with the agent's dialect_style setting to switch its delivery — for example, rendering English as Nigerian Pidgin.

Local dialects

Dialect-capable voices cover the regional varieties Aethex is tuned for — most prominently Nigerian Pidgin on English voices. A voice without supports_dialect_style only speaks its standard language.

Browse the catalog

Use GET /voices to list all voices on your account. Add ?language= to narrow by language.

List English voices
bash
curl "https://api.aethexai.com/api/v1/voices?language=english" \
  -H "X-API-Key: $AETHEX_API_KEY"

Filter by dialect support

Add ?supports_dialect_style=true to GET /voices to list only voices that can speak a local dialect. Combine it with ?language= to narrow further.

List dialect-capable English voices
bash
curl "https://api.aethexai.com/api/v1/voices?language=english&supports_dialect_style=true" \
  -H "X-API-Key: $AETHEX_API_KEY"

Preview a voice

Before committing to a voice, run a text-through-TTS preview. Send a short phrase and receive a WAV file back.

Generate a preview
bash
curl -X POST "https://api.aethexai.com/api/v1/voices/preview" \
  -H "X-API-Key: $AETHEX_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"voice_id": "5c34046a-ac9b-57d5-8c70-5a61e694be3f", "text": "Welcome to Aethex. How can I help you?"}' \
  --output preview.wav

Preview a local dialect

To hear how a dialect-capable voice renders a local variety, send a phrase written in that dialect to POST /voices/preview with a voice that reports supports_dialect_style: true.

Preview a Pidgin phrase
bash
curl -X POST "https://api.aethexai.com/api/v1/voices/preview" \
  -H "X-API-Key: $AETHEX_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"voice_id": "5c34046a-ac9b-57d5-8c70-5a61e694be3f", "text": "Welcome come to Aethex oh! Wetin I fit do for you?"}' \
  --output preview_pidgin.wav

Picking a voice

A few tips when choosing:- Match the persona. The voice name and tone should align with your agent's system_prompt persona. A formal banking agent pairs better with a measured voice; a casual support agent benefits from a warmer one.
- Test with real phrases. Run your opener and a few typical responses through POST /voices/preview before going live.
- Check for dialect consistency. If you set dialect_style, pick a voice trained on the same dialect.

Next

Once you have a voice, assign it: set voice_id on the agent with PATCH /agents/{id}. If you need the agent to speak multiple languages, create separate agents per language, each with a matching voice.