Everything people ask before their first call
Cost, languages, numbers, data and how the agent behaves on a real line. If your question is not here, write to us and we will answer it directly.
How much does an AI phone call cost on Aitelz Voice?
About Rs 2 per minute all in, covering the voice, speech recognition, the LLM brain and Indian telephony. A comparable ElevenLabs plus Twilio plus GPT stack typically lands at Rs 15 to 20 per minute. Billing is in INR. The full breakdown is on the pricing page.
Does it understand Hindi and Hinglish?
Yes. Speech recognition runs on Deepgram, which is Hinglish capable, so callers can mix Hindi and English mid sentence the way people actually talk on Indian phone lines. The agent replies in the style you set in its persona.
How long does setup take?
Building an agent takes minutes: write the persona and greeting in the agent builder, pick a voice in the voice studio, and test it from the console. Going live on a phone number then depends only on assigning a DID to the agent.
How does telephony work in India? Do I get real numbers?
Calls run on real Indian DIDs provisioned through Vobiz and VoiceLink. The platform handles both inbound calls to your numbers and outbound dialing, with routing, wallet and engine health visible live in the telephony console.
Where does my call data live? Can I self-host?
The platform is self hosted, so it runs on infrastructure you control and your call data, recordings and transcripts stay with you. There is no requirement to ship customer conversations to a third party SaaS.
Can callers interrupt the agent mid sentence?
Yes, that is barge-in. When a caller starts speaking, the agent stops talking and listens, exactly like a human would. Combined with sub 2 second responses, conversations feel natural rather than scripted.
How fast does the agent respond?
First audio typically lands in under two seconds. That latency is the difference between a call that feels like a person picking up and one that feels like a bot buffering, so we treat it as a core product metric.
Do I get recordings and transcripts?
Yes. Every call produces a recording you can replay and a transcript you can search. Teams use them for quality audits, dispute resolution, agent tuning and client reporting.
Which providers power the stack? Am I locked in?
Rumik silk TTS for the voice, about 20x cheaper than ElevenLabs. Deepgram for speech recognition, Groq Llama 3.3 for the brain, and Vobiz and VoiceLink for Indian telephony. Every layer is a swappable adapter, so changing a vendor is a setting, not a rewrite.
Can agencies white-label it for their clients?
Yes. The platform is multi tenant at the core: each client gets an isolated workspace with its own agents, numbers, branding and usage, and tenant isolation is enforced on every read and write. Agencies resell from one console under their own brand. Details on the agencies page.
Does it integrate with my existing tools?
The platform exposes a clean JSON API for agents, calls, usage and transcripts, so you can wire bookings and leads into your CRM, calendar or sheet. Because it is self hosted and provider agnostic, there is no walled garden between you and your own data.
Is there a trial? How do I start?
Open the console, sign up, build an agent and hear it speak in the voice studio before any phone spend. For a guided walkthrough or agency setup, write to hello@brandingpioneers.com.