Skip to content
TriedDocs
Build agents

Voice

Pick a voice and a voice tier, Premium or Standard, and set voice temperature, speed and volume.

The voice is how the agent sounds. You make two choices: which voice, and which tier. Premium is the default and sounds the most natural. Standard is clear and fast at half the price. Pick both with the voice button in the toolbar, which shows the voice name and tier.

The Select Voice dialog with the Premium and Standard tiers and the voice list

Select a voice#

Open the picker#

Click the voice button in the toolbar. Select Voice opens.

Choose a tier#

Pick Premium (recommended) or Standard at the top. Each card shows the price per minute.

Choose a voice#

Click a voice card. Each card shows the voice's gender and age, and has a play button for a sample. Your workspace's cloned voices appear in the same list.

Save#

Click Use voice. This sets voice_id and voice_model on the agent.

Clone a voice in the footer adds your own voice. See Voice cloning.

Voice tiers#

Tier voice_model Character Estimated first audio Voice $/min
Premium (default) eleven_v4_turbo Most natural and expressive. Performs expressive tags. ~190 ms 0.08
Standard eleven_flash_v2_5 Clear and fast ~180 ms 0.04

The tier shows on every call's cost breakdown as Voice (Premium) or Voice (Standard), and as the voice_premium or voice_standard line in billing usage.

With the default model, GPT-5.6 Terra, a call costs $0.30 per minute (AUD) with Premium and $0.24 with Standard. See Billing.

Voices#

Voice voice_id
Adrian (default) 11labs-Adrian
Rachel 11labs-Rachel
Sarah 11labs-Sarah
Drew 11labs-Drew
Matilda 11labs-Matilda
Antoni 11labs-Antoni
Lily 11labs-Lily
Dave 11labs-Dave

Every voice works on both tiers. List voices returns these plus your workspace's cloned voices, each with gender, accent and age.

Needs keys Real voice audio needs the platform's voice key. Without it, previews use the browser's own speech engine, which only approximates the voice, and calls run on the mock voice.

Other voice models through the API#

For agents moved from Retell, the API also accepts the other voice_id and voice_model values that List voices returns. The dashboard shows these as the voice name only. New agents should use one of the two tiers above.

Voice settings#

These sliders are under Speech Settings. See Speech settings for the rest of that section.

Setting Field Range Default Effect
Voice Speed voice_speed 0.5 to 2 1 Speaking rate.
Voice Temperature voice_temperature 0 to 2 1 Higher is more expressive but less stable.
Volume volume 0 to 2 1 Output loudness, applied to the audio.

How speed is applied#

The voice engine accepts speeds from 0.7 to 1.2. Tried sends the part of voice_speed the engine supports and applies the rest with a time-stretch that keeps the pitch, so the full 0.5 to 2 range works on both tiers.

Dynamic voice speed (enable_dynamic_voice_speed) adjusts the speed during the call to match the caller's pace, and responds when the caller asks the agent to speak faster or slower. See Speech settings.

How temperature is applied#

voice_temperature sets the voice's stability: temperature 1 gives stability 0.5, temperature 0 gives the most stable delivery, and 2 the least stable.

Set the voice with the API#

curl -X PATCH https://api.tried.com.au/update-agent/agent_3f1c9a7b2d4e6f8091a2b3c4 \
  -H "Authorization: Bearer $TRIED_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{ "voice_id": "11labs-Matilda", "voice_model": "eleven_flash_v2_5", "voice_speed": 1.1, "voice_temperature": 0.8 }'

This switches the agent to Matilda on the Standard tier.