Configure ElevenLabs Voices & Models
Pick the right voice and model for your use case, and tune stability if needed.
4 min readUpdated 2026-02-19
ElevenLabs powers Voxinity's voice synthesis. Each agent picks one voice and one model. Voxinity handles streaming, latency optimization, and barge-in detection automatically.
Picking a voice
- Open the agent editor → Voice tab → Browse Voices.
- Filter by language, age, gender, and use-case (conversational, narration, calm, energetic).
- Hit the play icon next to any voice to preview a sample phrase using your agent's tone.
- Click Use Voice to assign it.
Model trade-offs
Models
eleven_flash_v2_5
Lowest latency (~250ms), broad voice support, best for conversational sales/support.
eleven_turbo_v2_5
Slightly higher quality than Flash, same latency tier.
eleven_multilingual_v2 / v3
Premium quality, multi-language, ~100ms higher latency. Use for hospitality, premium brands.
Stability, similarity, style
- Stability — higher = more consistent emotion, less variability. We default to 0.5.
- Similarity Boost — how close to the original voice clone the synthesis stays. Default 0.75.
- Style — exaggeration of the voice's natural style. Keep at 0 unless you want theatrical delivery.
Cloning
Voice cloning is available through your own ElevenLabs Pro account. Once a clone is created on ElevenLabs, it appears in Voxinity's voice picker automatically.
← Previous
Import Phone Numbers from Twilio or Telnyx
Next →
Interaction Memory — One Brain Across Every Channel
Still have questions?
Reach our team — replies within 1 business day.