Voice Engines
Pick an engine, then the models it allows. Every price is per minute.
Filters
Good for
Languages
Cost per minute
Models by
7 results
Gemini Live$0.0491One model for the whole turn.models by Google · Speech-to-speech · 97 languages · 12 Indian languages
1 model on this engine
Speech-to-speech model
Sarvam$0.0617Indian languages end to end.models by Sarvam · Pipeline · 11 languages · 10 Indian languages · from $0.0343/min5 models on this engine
Speech-to-text
- Saaras v3 Realtime+$0.0076/min
- Saaras v4+$0.0076/min
Language model
- Sarvam-105B+$0.0010/min
Voice
- Bulbul v3+$0.0411/min
- Bulbul v4 Flash+$0.0137/min
OpenAI GPT-Live$0.0827OpenAI GPT-Live handles the voice: it listens and speaks at the same time and manages turn-taking.models by OpenAI · Speech-to-speech · 32 languages · 2 Indian languages · Any of 32 LLMs · from $0.0823/min
33 models on this engine
Speech-to-speech model
Language model
Claude Haiku 4.5+$0.0067/min
DeepSeek V3.1+$0.0015/min
Mercury 3+$0.0006/min
Llama 3.3 70B+$0.0006/min
Mistral Medium 3.5+$0.0101/min- Mistral Small 26.03+$0.0009/min
Nemotron 3.5 Lightning+$0.0004/min
Qwen3.7 Flash+$0.0002/min
Grok 4.3+$0.0059/min
MiMo V2.6 Flash+$0.0007/min
GLM 5.3 Flash+$0.0008/min
2 models on this engine
Speech-to-speech model
Classic$0.0865The default pipeline for English.models by Deepgram, ElevenLabs, Soniox, Microsoft, Cartesia, Google, Inworld · Pipeline · 42 languages · 12 Indian languages · Any of 32 LLMs · from $0.0293/min51 models on this engine
Speech-to-text
- Nova-3+$0.0108/min
- Nova-3 Multilingual+$0.0129/min
Scribe v2 Realtime+$0.0091/min
Real-time v5+$0.0028/min
Language model
- Mistral Medium 3.5+$0.0101/min
- Mistral Small 26.03+$0.0009/min
Voice
Azure Neural voices+$0.0189/min
Sonic 3.6+$0.0819/minTTS 2+$0.0315/min
- TTS v2+$0.0163/min
1 model on this engine
Speech-to-speech model
- Deepgram Agent$0.1851Deepgram's Voice Agent API runs the conversation as a single realtime session and drives ElevenLabs Turbo v2.5 for the voice itself.models by Deepgram, ElevenLabs · Speech-to-speech · 32 languages · 2 Indian languages
4 models on this engine
Speech-to-speech model
- Voice Agent$0.1851
Voice
A pipeline engine's minute is the models you pick plus the $0.012 per minute platform fee. Pick them and see the total in the pipeline builder.
Questions about voice engines
What is the cheapest voice engine for an AI phone agent?
Gemini Live is the cheapest engine at $0.0491 per minute, platform fee included. Classic with its cheapest models comes to $0.0293 per minute. Every price here is the one the meter bills, per second.
Which voice engine is best for callers in India?
Gemini Live speaks 12 Indian languages in one model, at $0.0491 per minute. The Sarvam engine is built for Indian languages end to end. On Classic, ElevenLabs v3 or v4 voices with Scribe v2 Realtime cover all twelve Indian languages listed here.
Should I use a speech-to-speech engine or a pipeline engine?
A speech-to-speech engine listens, thinks and speaks in one model, so it is the quickest to set up. A pipeline engine lets you pick each part: the speech-to-text, the voice, and on Classic any of 32 language models. GPT-Live is both: one voice model, with the language model of your choice behind it.
Can I switch an agent to another engine or model later?
Yes. The engine and its models are settings on the agent, changed in the console or with one API request, with no code to rewrite. New engines and models are added to Wixzel Voice as they ship.
Do I need an account with each model provider?
No. Every engine and model here runs on your Wixzel Voice API key and one prepaid balance, so there are no separate provider accounts, keys or bills.
How are engine prices calculated?
An engine's price is a typical connected minute with its default models, platform fee included. The meter bills what actually ran, per second: seconds of audio, tokens and characters at each model's own rate. A model's figure under an engine is its share of a typical minute.
Build your first voice agent