OpenAI Realtime API alternative

Speech to speech, on Indian phone calls.

OpenAI sells tokens. We sell the whole call: speech, model, voice and the phone line, in one rupee price.

OpenAI’s numbers checked 25 October 2026.

AI agents from ₹1 a minute.¹

Compare
The Frequency console: Agent Studio with four agents in Hinglish, Hindi, Tamil and Gujarati, and a test call in the browser.

Why switch

Why people switch to Frequency.

Everything a phone call in India needs, from one team, priced in rupees.

AI agents

AI agents from ₹1 a minute.¹

Speech recognition, the model and the voice, in one rupee rate. Billed in 10-second blocks, so short calls stay cheap.

Choose your plan

Five languages, live.

Hindi, English, Gujarati, Tamil and Telugu. And the Hinglish in between, mid-sentence.

Every call runs on our own GPUs in India.

The listening, the thinking and the voice are ours: Fourier STT, Gemma 4 26B and Fourier TTS.

One bill, in rupees.

The phone line, the platform, speech recognition, the model and the voice all come from us.

Live the same day.

Build in the sandbox, press Go live, and call today. Every company starts on pay as you go, with no monthly commitment.

Hear it before you decide.

Real lines from Fourier TTS, one request each, unedited. Pick a language and press play.

Hear it

OpenAI sells the model. You build the call.

Here is how OpenAI and Frequency differ.

The Realtime API is a model you build a phone agent around: one socket that listens and speaks, priced per audio token, with a SIP trunk and a server of your own. We sell the finished call. Hindi, Gujarati, Tamil, Telugu and English agents on our own stack in India, one price a minute in rupees, and the phone line inside it on pay as you go.

The Realtime API is a brilliant piece of a phone agent. One socket in, speech out, with tool calls in between. What it is not is a phone agent. You still need a server to run each call, a SIP trunk to reach a phone, numbers from a carrier, a dialler, retries, recordings and a webhook into your system. OpenAI’s own SIP guide says to use a trunking provider such as Twilio.

We sell the finished thing. The speech recognition, the model and the voice are our own, served on our own GPUs in India. The model is the open Gemma 4 26B. On pay as you go, the phone line and +91 numbers are inside the rate, and the result of every call comes back to you as a signed webhook.

Hear it.

Real lines from Fourier TTS. Pick one, press play.

Text149 characters

Hi, मैं नेहा बोल रही हूँ Acme Store से। आपका order dispatch हो गया है और Friday तक deliver हो जाएगा। Address में कोई change करना हो तो अभी बता दीजिए।

48 kHz · MonoReady when you are

0:00 / 0:09

voice
Neha
language
Hinglish
speech
0:09
characters
149

Fourier TTS, voice neha, rendered 16 Oct 2026. One request, not edited.

Text119 characters

नमस्ते, मैं श्रुति बोल रही हूँ, दाँतों के क्लिनिक से। कल सुबह डॉक्टर साहब से आपका मिलना तय है। क्या आप समय पर आ पाएँगे?

48 kHz · MonoReady when you are

0:00 / 0:08

voice
Shruti
language
Hindi
speech
0:08
characters
119

Fourier TTS, voice shruti, rendered 16 Oct 2026. One request, not edited.

Text117 characters

નમસ્તે, હું આશુ બોલું છું. તમારી લોનનો હપ્તો આવતા અઠવાડિયે ભરવાનો છે. તમે કહો તો હું અત્યારે જ ભરવાની રીત સમજાવી દઉં.

48 kHz · MonoReady when you are

0:00 / 0:08

voice
Ashu
language
Gujarati
speech
0:08
characters
117

Fourier TTS, voice ashu, rendered 17 Oct 2026. One request, not edited.

Text167 characters

வணக்கம், நான் நேஹா பேசுகிறேன். உங்கள் பார்சல் நாளை மாலைக்குள் வந்து சேரும். அந்த நேரத்தில் நீங்கள் வீட்டில் இல்லையென்றால் சொல்லுங்கள், வேறு நாளுக்கு மாற்றி விடுகிறோம்.

48 kHz · MonoReady when you are

0:00 / 0:10

voice
Neha
language
Tamil
speech
0:10
characters
167

Fourier TTS, voice neha, rendered 17 Oct 2026. One request, not edited.

Text114 characters

నమస్కారం, నేను సమీర్ మాట్లాడుతున్నాను. నిన్న మీరు మా కొత్త ఇళ్ళ గురించి అడిగారు. ఇప్పుడు కొంచెం సేపు మాట్లాడవచ్చా?

48 kHz · MonoReady when you are

0:00 / 0:08

voice
Sameer
language
Telugu
speech
0:08
characters
114

Fourier TTS, voice sameer, rendered 17 Oct 2026. One request, not edited.

Side by side

Here is OpenAI beside Frequency.

Every price in rupees², from their own pages and ours.³

Choose your plan.

Choose your plan.

Your price · T1

₹1 a minute¹

₹1,000 for 1,000 minutes

₹1,00,000 / month minimum, credited to your wallet; your usage counts towards it. Our phone line adds ₹0.30 and call insights ₹0.20 a minute, if you want them. OpenAI publishes no volume price, so the chart shows ours beside its notes.

Your price · T3

₹3 a minute¹

₹3,000 for 1,000 minutes

₹50,000 / month minimum, credited to your wallet; your usage counts towards it. The phone line and call insights are in the rate. OpenAI publishes no volume price, so the chart shows ours beside its notes.

Your price · T5

₹5 a minute¹

₹5,000 for 1,000 minutes

Pay as you go from your wallet. The phone line and call insights are in the rate. In the chart, beside OpenAI’s pay-as-you-go price.

One minute of a call, in rupeesYour plan, beside the price it pairs with
One minute of a call, OpenAI Realtime API and Frequency, in rupees
OptionBar₹ a minute
Pay as you goGPT-Live's voice session, the one OpenAI voice price printed per minute, against our T5
OpenAI GPT-Live the voice session on its own + the backend model's tokens, and a SIP carrier's line ₹4.75$0.05
Frequency T5 pay as you go, no commitment ₹5all-in
At volumeOpenAI prints no volume rate for voice; the gpt-realtime models are priced per audio token, and we do not turn a token price into a minute
Frequency T3 ₹50,000/month commitment ₹3all-in
Frequency T1 ₹1,00,000/month commitment + ₹0.30 phone line and ₹0.20 insights, if you want them ₹1₹1.50 with both
How we read these numbers

Dollar prices converted at ₹94.96, the ECB reference rate on 16 October 2026. The rate.

Each bar is one minute as OpenAI’s own pages price it, layer by layer, beside ours from our price list. A hatched end is a part of the call their price leaves out and publishes no number for, so it has no length.

  • OpenAI GPT-Live: voice session ₹4.75 ($0.05 per minute)

GPT-Live is the one OpenAI voice price printed per minute: $0.05 per minute, which is ₹4.75 at the ECB rate, billed per second, before the backend model and the phone line. The gpt-realtime models are priced per audio token, $32.00 in and $64.00 out per 1M tokens for gpt-realtime-2.1, and because every turn resends the whole conversation, a minute's cost depends on how long the call has already run. So that bar has no length.

From developers.openai.com/api/docs/models/gpt-live-1, checked . Ours from our price list.

The chart shows the plan you choose above beside the price it pairs with: our T1 and T3 beside OpenAI’s volume price, our T5 beside its pay-as-you-go price. Without that choice, every bar is listed.

What a minute costs, OpenAI Realtime API and Frequency, in their words
DetailOpenAI Realtime APIFrequency
Published price per minute GPT-Live voice sessions are $0.05 per minute, billed per second, with the backend model and tool usage billed separately the gpt-realtime models publish no per-minute price, only token rates developers.openai.com T1 ₹1/min with ₹1,00,000/month T3 ₹3/min with ₹50,000/month; T5 ₹5/min pay as you go, phone line and insights included (GST extra)
Audio tokens gpt-realtime-2.1 audio: $32.00 input, $64.00 output and $0.40 cached input per 1M tokens gpt-realtime-2.1-mini audio: $10.00 input and $20.00 output; a caller's audio is 1 token per 100 ms of audio and the assistant's is 1 token per 50ms of audio developers.openai.com No tokens to count: the whole call is one rate a minute, in 10-second blocks
What the price includes The voice model's tokens, or the GPT-Live session the backend model, tools, the SIP trunk and the carrier's minutes are separate, and every turn resends the whole conversation, so turns later in the session will be more expensive developers.openai.com The phone line, speech recognition, the model, the voice and call insights on T5 and T3 on T1 the line and insights are optional at ₹0.30/min and ₹0.20/min

What it costs

OpenAI prints one voice price per minute, and it is for GPT-Live: $0.05 per minute, billed per second. That is ₹4.75 at the ECB rate, for the voice session alone. The backend model’s tokens, your server and the carrier’s minutes come on top.

The gpt-realtime models have no per-minute price at all. They are priced per audio token, $32.00 in and $64.00 out per 1M tokens for gpt-realtime-2.1, and because every turn resends the whole conversation, the tenth minute of a call costs more than the first. OpenAI says as much in its cost guide.

Ours is ₹5 a minute on pay as you go, with the line inside. A 1,000-minute month is ₹5,000. Once you call every day, the volume plan takes it to ₹1 a minute. Every minute costs the same, in 10-second blocks, and if nobody picks up, you pay nothing.

Who runs each part of the call

OpenAI sells the model that listens and speaks. The phone line, the server that runs the call and anything India-side are yours to arrange. We are the models and the platform.

Who runs each part of the call, OpenAI Realtime API and Frequency
DetailOpenAI Realtime APIFrequency
Phone line You bring it A SIP trunk, such as Twilio developers.openai.com Ours Mirai phone line, +91 numbers
Platform Theirs GPT-Live, or your own server developers.openai.com Ours Our orchestration
Speech recognition Theirs Inside the speech-to-speech model developers.openai.com Ours Fourier STT
Language model Theirs gpt-realtime-2.1, per token developers.openai.com Ours Gemma 4 26B, on our GPUs
Voice Theirs Inside the speech-to-speech model developers.openai.com Ours Fourier TTS

Languages

Languages and voices, OpenAI Realtime API and Frequency
DetailOpenAI Realtime APIFrequency
Languages No supported-language list is published for the Realtime API or GPT-Live the prompting guide names Hindi only as an accent of English; ten voices: alloy, ash, ballad, coral, echo, sage, shimmer, verse, marin and cedar developers.openai.com Hindi, Gujarati, English, Tamil and Telugu, plus Hinglish, in speech recognition and voice Bengali, Marathi and Kannada are coming

Where it runs

Where it runs, OpenAI Realtime API and Frequency
DetailOpenAI Realtime APIFrequency
Where it runs India (in.api.openai.com) is listed for regional storage, not regional processing, and the Realtime API is not among the endpoints listed for it OpenAI says support for regional storage does not imply support for regional processing developers.openai.com Every call runs on our own GPUs in India

Phone numbers

Phone numbers, OpenAI Realtime API and Frequency
DetailOpenAI Realtime APIFrequency
Phone numbers None of its own: calls arrive over SIP from a trunking provider such as Twilio, and outbound SIP calling must be enabled for your organisation ringing is limited to 3 minutes, a connected call to 2 hours, and a Realtime session to 60 minutes developers.openai.com Indian (+91) numbers: buy a dedicated number through us, or connect your own Plivo number

Transfer to a person

More, OpenAI Realtime API and Frequency
DetailOpenAI Realtime APIFrequency
Transfer to a person Yes, with a SIP REFER to a target of your choosing developers.openai.com Not yet; the agent can promise a callback, and your system picks it up from the call's outcome

Spot something out of date? Tell us. Six platforms in one table: Voice AI platforms compared.

Which to pick

Which is right for you?

Sometimes it's them. We'll say so.

Choose OpenAI if…

  • You want to build the whole agent yourself around one speech-to-speech model, with your own server and your own SIP trunk.
  • Your callers are outside India, or mostly speak English.
  • You need a published certifications list today: SOC 2 Type 2, ISO 27001 and more.
  • You need to transfer a live call to a person over SIP today.

Choose Frequency if…

  • Your customers are in India and speak Hindi, Gujarati, Tamil, Telugu, English or a mix.
  • You want one price per minute, in rupees, instead of tokens in dollars.
  • You want the phone line and +91 numbers inside the product, not from a third trunk.
  • You make a lot of calls. Our agents cost ₹1 a minute on the volume plan.

Where OpenAI is ahead

A few things to know before you move. The Realtime API can transfer a live call to a person with a SIP REFER, which we cannot do yet. Its certifications list is long and public. It takes WebRTC straight from a browser, and GPT-Live listens while it speaks. And it is a general model: if your callers are mostly in English, it will do well.

What it does not publish is a language list. Hindi appears in its prompting guide only as an accent of English. India is on its data-residency list for storage, not processing, and the Realtime API is not among the endpoints listed for the India region. The latency guide shows how to time the two on real calls to Indian phones.

Moving

Moving your agent across is easy.

Bring the agent you have. Most of it carries across.

  1. Step 1

    Open the sandbox.

    Look around with no sign-up. We ask for an email only when you test a call.

  2. Step 2

    Bring your prompt.

    Paste the prompt you already wrote into Agent Studio, and pick a voice and a language.

  3. Step 3

    Talk to it. Go live.

    Test it in your browser or on your own phone, then go live the same day.

Open the sandbox

Moving a Realtime agent

Your system prompt and first message become system_prompt and first_message on a new agent. Your tool calls become the agent’s tools, and the result arrives on your webhook when the call ends. You start a call with POST /v2/calls to a +91 number you bought from us or brought from Plivo. The quickstart goes from a new key to a real phone call in five minutes.

If you only want an Indian voice inside the agent you already built, Fourier TTS speaks OpenAI’s speech API: change the base URL and the model name, and keep the rest. Then put both on your own phone with your real script, and listen.

Questions

Here is what people ask before switching from OpenAI.

Is $0.05 a minute comparable to your ₹5?

No. GPT-Live's $0.05 per minute, about ₹4.75, is the voice session alone. The backend model's tokens, your server and the SIP carrier's minutes come on top. Our ₹5 on pay as you go is the whole call, phone line included.

What does the Realtime API cost a minute?

OpenAI does not say. The gpt-realtime models are priced per audio token, $32.00 in and $64.00 out per 1M tokens, and each turn resends the conversation so far. A long call costs more per minute near the end than at the start. We bill every minute the same, in 10-second blocks.

Can the Realtime API speak Hindi?

OpenAI publishes no language list for it, and its prompting guide mentions Hindi only as an accent of English. Test it on your own script before you decide. Hindi, Gujarati, Tamil, Telugu and English are live in our speech recognition and our voice, and the agent slips into English mid-sentence the way callers do.

What does a 1,000-minute month cost?

On GPT-Live, 1,000 minutes of voice sessions is $50, about ₹4,748 at the ECB rate, before the backend model and the phone line. On our pay-as-you-go plan the same 1,000 minutes cost ₹5,000 with the line included, and ₹1 a minute on the volume plan.

Can I keep my OpenAI code?

For the voice alone, yes. Fourier TTS takes OpenAI's speech request, so you change the base URL, the model and the voice. For a whole phone agent, the Realtime socket does not carry across: you bring the prompt and the first message to a new agent, and we run the call.

Still deciding? Ask us.

Fifteen minutes with the team that builds it. Bring your script, or your hardest call.

Book 15 minutes

Related

Book a call

15 minutes with Sneh Mehta

Finding a time for us…

Powered by Cal.comOpen calendar in a new tab ↗

  1. AI agents from ₹1 a minute: on our volume plan, ₹1,00,000 a month. Pay as you go is ₹5 a minute. Prices exclude GST.

  2. Dollar prices converted at ₹94.96, the ECB reference rate on 16 October 2026. The rate.

  3. OpenAI Realtime API’s figures are from its own pages, linked beside each one and listed under Sources, checked . Ours are from our price list and docs.