Voices: catalog, languages, and how to pick one
Every voice available on Call2Me — providers (ElevenLabs, OpenAI, Google), languages, gender, and how to audition them before going live.
Updated May 6, 2026
A voice is what your agent sounds like. Picking the right one is half brand decision, half latency-and-quality trade-off. This is the catalog and the decision framework.
Provider overview
| Provider | Best for | Notable trade-off |
|---|---|---|
| OpenAI | Lowest latency, neutral voices | Smaller catalog, fewer languages |
| ElevenLabs | Most expressive, voice cloning, multilingual | Slightly higher latency, premium tier required for cloning |
| Google (Wavenet) | Strong language coverage, predictable cost | Less natural prosody than ElevenLabs |
The platform negotiates the right provider per voice — you just pick a
voice_id.
Languages
| Code | Language | Voices available |
|---|---|---|
en-US | English (US) | All providers |
tr-TR | Turkish | ElevenLabs, Google |
de-DE | German | All providers |
fr-FR | French | All providers |
es-ES | Spanish | All providers |
it-IT | Italian | ElevenLabs, OpenAI |
pt-BR | Portuguese (BR) | ElevenLabs, OpenAI |
ar-XA | Arabic | ElevenLabs, Google |
multilingual | Auto-detect across the above | ElevenLabs Multilingual v2 |
For a single-language agent, set language to the locale and pick a
voice in that language. For a multilingual agent, pick
elevenlabs-multilingual-v2 and set language: "multilingual".
Auditioning
The fastest way to pick: dashboard → Agents → Voice → Gallery. Each entry has a play button — you hear a 3-second sample in the language/gender/style you want.
Programmatically:
curl https://api.call2me.app/v1/voices?language=tr-TR \
-H "Authorization: Bearer YOUR_API_KEY"
Returns the catalog filtered by language. Each entry has a
preview_audio_url you can play.
Picking a voice — the framework
Three questions, in order:
1. Language coverage
If your callers speak one language: any voice in that language will do. If multiple: pick a multilingual voice unless brand consistency matters more than character consistency.
2. Brand vs. utility
Some voices feel premium (warmer, more expressive); some feel utility (neutral, fast). For a luxury hotel: pick expressive. For a transactional support line: pick fast and neutral — your callers want the answer, not the performance.
3. Latency budget
If your TTS budget is < 200ms, OpenAI is the safe default. If you can afford 250ms, ElevenLabs sounds noticeably more natural and is worth it for most consumer-facing agents. See Voice for the full latency breakdown.
Voice cloning
For your own voice (CEO's voice, brand voice talent), clone via ElevenLabs:
- Sign up at elevenlabs.io
- Use their voice cloning tools to create your custom voice
- Note the voice ID they give you (looks like
21m00Tcm4TlvDq8ikWAM) - In Call2Me, set the agent's
voice_idtoelevenlabs-<their-id>
ElevenLabs charges separately for cloned-voice usage at their published rates. The platform passes calls through; we don't mark up.
Per-language voice consistency
If your agent runs in multiple languages on a single call (rare but
happens), pick a multilingual voice so the character stays
consistent. Switching from en-US/nova to tr-TR/aysel mid-call
breaks immersion; ElevenLabs Multilingual v2 stays the same character
across all 9 languages.
What's next
Frequently asked
Q.How many voices are available?
Roughly 40+ across providers, covering English, Turkish, German, French, Spanish, Italian, Portuguese, Arabic and a multilingual class. The dashboard's voice gallery has a preview button on every entry.
Q.Can I clone my own voice?
Yes via ElevenLabs — sign up there, clone a voice with their tools, then set the agent's voice_id to your cloned voice. The platform proxies the synthesis call. ElevenLabs charges separately for cloned-voice usage.
Q.Which voice is fastest?
OpenAI voices have the lowest first-token latency (typically <200ms). ElevenLabs is more expressive but ~50–100ms slower. Google Wavenet is in between. Pick fastest when the conversation is transactional; pick most expressive when the brand voice matters.
Q.Can a single agent speak multiple languages?
Yes. Pick a multilingual voice (ElevenLabs Multilingual v2 is the most natural across our 9 languages) and the agent will adapt to whatever language the caller speaks.