OpenAI speech-to-text alternative

Transcripts of Indian calls, with every speaker labelled.

One request in, a transcript out with a speaker on every line. Our voice speaks OpenAI’s speech API on the same key.

OpenAI’s numbers checked 25 October 2026.

Get an API key Compare
OpenAI speech-to-text and Voice Infra at a glance
OpenAI speech-to-text Voice Infra
Transcription price
₹25.64an hour of audio, gpt-transcribe ($0.0045 / minute) ₹15an hour of audio
Indian languages
On Whisper's list; none printed for gpt-transcribe Hindi and seven more, plus Hinglish
Speaker labels
Only on a deprecated model Included
Whisper
Removed on February 26, 2027 One model, no retirement date
Every detail, with sources

Side by side

Here is OpenAI beside Voice Infra.

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

Price

Recorded calls, ₹ per hour of audio
Recorded calls, OpenAI speech-to-text and Voice Infra, in rupees per hour of audio
OptionBar₹ per hour of audio
gpt-transcribe the current model ₹25.64$0.27
gpt-4o-transcribe deprecated, until February 26, 2027 ₹34.19$0.36
gpt-live-transcribe live, no speaker labels ₹96.86$1.02
Mira Transcribe recorded calls, speaker labels included ₹15
How we read these numbers

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

OpenAI prices by the minute, so an hour is 60 of them: $0.0045 × 60 = $0.27, $0.006 × 60 = $0.36 and $0.017 × 60 = $1.02. Speaker labels are inside our ₹15; on OpenAI they come only from a model that is being retired.

From developers.openai.com/api/docs/pricing, checked . Ours from our price list.

What speech costs, OpenAI speech-to-text and Voice Infra, in their words
DetailOpenAI speech-to-textVoice Infra
Transcription price gpt-transcribe $0.0045 / minute the deprecated gpt-4o-transcribe $0.006 / minute and gpt-4o-mini-transcribe $0.003 / minute until February 26, 2027 developers.openai.com ₹15 per hour of audio (₹0.25 a minute), charged once when the job completes speaker labels included
Live transcription price gpt-live-transcribe $0.017 / minute and gpt-realtime-whisper $0.017 / minute, over a Realtime session developers.openai.com Streaming recognition runs inside every voice agent as a standalone API it is in private preview, on request

What an hour costs

gpt-transcribe is $0.0045 a minute, which is $0.27 an hour, about ₹25.64 at the ECB rate. The deprecated gpt-4o-mini-transcribe is cheaper, about ₹17.09 an hour, until it goes. Live transcription on gpt-live-transcribe is about ₹96.86 an hour. None of those include speaker labels.

Ours is ₹15 an hour, 25 paise a minute, speaker labels included. A five-minute call costs ₹1.25. You pay once, when the job finishes, and a failed or degraded job is free.

The languages that matter here

Whisper’s list has 98 languages, and Hindi, Tamil, Telugu, Gujarati and Marathi are all on it. OpenAI’s own note says accuracy varies by language, and it publishes no number per language. For gpt-transcribe there is no list at all.

We publish ours. On 499 public Hindi phone recordings at 8 kHz, Mira STT made 18.3% word errors, against 21.6% for Sarvam’s Saaras v4 on the same clips. On clean read speech Saaras does better, 5.8% against our 10.1%. Send twenty of your own recordings and score them.

Indian languages

Languages and voices, OpenAI speech-to-text and Voice Infra
DetailOpenAI speech-to-textVoice Infra
Indian languages Whisper's own list has 98 languages, Hindi, Tamil, Telugu, Gujarati and Marathi among them, and OpenAI says accuracy varies by language the docs print no language list for gpt-transcribe, which takes ISO language hints developers.openai.com Hindi, Gujarati, Marathi, Bengali, Tamil, Telugu, Kannada and Malayalam, plus Hinglish Hindi in Devanagari, English in Latin script

Where it runs

Where it runs, OpenAI speech-to-text and Voice Infra
DetailOpenAI speech-to-textVoice Infra
Where it runs India (in.api.openai.com) is on the data residency list for storage, not processing, and requires MAM or ZDR approval residency endpoints carry a 10% uplift for newer models developers.openai.com Speech recognition runs on our own GPUs in India

How to switch

Getting started, OpenAI speech-to-text and Voice Infra
DetailOpenAI speech-to-textVoice Infra
How to switch POST /v1/audio/transcriptions with the file, and stream=true for partial text as it processes developers.openai.com POST /v2/transcriptions with the file, a line of context and the names to get right for speech, the same OpenAI POST /v1/audio/speech protocol with model mira-tts

More

More, OpenAI speech-to-text and Voice Infra
DetailOpenAI speech-to-textVoice Infra
Whisper whisper-1 is $0.006 a minute on its model page and deprecated, with removal from the API on February 26, 2027 OpenAI points to gpt-live-transcribe or gpt-transcribe instead developers.openai.com One model, Mira Transcribe, at ₹15 an hour
Speaker labels Only on the deprecated gpt-4o-transcribe-diarize, until February 26, 2027 gpt-live-transcribe doesn’t return word-level timestamps, speaker labels, or transcription confidence scores developers.openai.com Included on every transcript, with a speaker id per voice
File limits Files can be up to 25 MB mp3, mp4, mpeg, mpga, m4a, wav and webm developers.openai.com Recordings up to 2 hours files over 45 MB by https link

Spot something out of date? Tell us. Every platform in one table: voice AI platforms compared.

Why switch

Why people switch to Voice Infra.

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

Mira Transcribe

₹15 an hour.⁴

Recorded calls, speaker labels included, charged once, when the transcript is done.

Eight Indian languages, live.

Hindi, Gujarati, Marathi, Bengali, Tamil, Telugu, Kannada and Malayalam. And the Hinglish in between, mid-sentence.

Speech recognition and the model run on our own GPUs in India.

The listening and the thinking: Mira STT and Gemma 4 26B. Mira TTS speaks.

Who said what.

Every transcript comes with speaker labels, and every word in its own script.

Phone agents, on the same key.

When you want the whole call, the same key and wallet run our voice agents.

AI agents: ₹5/min to start · ₹1.50/min all-in at volume.¹

Hear it before you decide.

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

Hear it

Cheap, everywhere, and being replaced

Here is how OpenAI and Voice Infra differ.

OpenAI's transcription API is cheap and everywhere: gpt-transcribe costs less than half a cent a minute, and most SDKs already speak it. But Whisper and the gpt-4o transcription models are being retired, the live model returns no speaker labels, and India is a storage region, not a processing one. We transcribe Indian phone calls at ₹15 an hour, with speaker labels included, on our own GPUs in India.

If you have ever sent audio to an API, there is a good chance it was Whisper. It is in every SDK and every tutorial, and it is cheap. OpenAI’s current model, gpt-transcribe, is cheaper still at less than half a cent a minute.

Two things have changed. OpenAI is retiring whisper-1 and the gpt-4o transcription models on February 26, 2027, including the only one that labels speakers. And the docs for gpt-transcribe print no list of supported languages, only that it takes language hints and that accuracy varies.

We do one thing: recorded phone calls in India. Mira Transcribe is our own speech recognition. It goes over the audio several times and our own language model writes one clean transcript. Give it a line of context and the names that matter, and it uses them. The result says who spoke, with Hindi in Devanagari and English words in Latin script.

Hear it.

Real lines from Mira 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

Mira 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

Mira 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

Mira 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

Mira 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

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

Which to pick

Which is right for you?

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

Choose OpenAI if…

  • Price per hour is what you are optimising. gpt-transcribe is about ₹25.64 an hour.
  • Your audio is mostly English, or spans dozens of languages.
  • You already run on OpenAI and want one bill for the model, the voice and the transcripts.
  • You need live transcription through a public API today, without speaker labels.

Choose Voice Infra if…

  • Your recordings are phone calls in Hindi, Gujarati, Marathi, Bengali, Tamil, Telugu, Kannada or Malayalam, often mixed.
  • You want speaker labels on every transcript, from a model that is not being retired.
  • You want the audio processed in India, on our own GPUs, with no approval to apply for.
  • You'd like an Indian voice that speaks OpenAI's own speech API, on the same key.

Where OpenAI is ahead

Price per hour on gpt-transcribe, by a clear margin. A live transcription API you can use today. One account for the language model, the voice and the transcripts. Dozens of languages, if your audio is not Indian. If you only need rough English transcripts of meetings, stay.

Moving

Moving your code across is easy.

One request in, a transcript with speakers out.

  1. Step 1

    Get an API key.

    Look around with no sign-up. When you are ready, create a key in Developers.

  2. Step 2

    Send a recording.

    One POST with the file, and a line of context about the call if you have it.

  3. Step 3

    Read who said what.

    The transcript comes back with a speaker on every line, each word in its own script.

Get an API key

Switching

Send the recording to POST /v2/transcriptions with a line of context and the names to get right. For files over 45 MB, send an https link instead. Then poll the job until it is done. The transcription guide has Python and Node.js examples.

Going the other way is even easier. Mira TTS speaks OpenAI’s speech API. Change the base URL, the model and the voice, and the same code that called OpenAI now speaks Hindi, Gujarati, Marathi, Bengali, Tamil, Telugu, Kannada or Malayalam on an Indian phone line. The setup guide shows the three lines.

POST /v2/transcriptions 
curl -X POST https://sandbox.voice.miraiminds.co/v2/transcriptions \
  -H "Authorization: Bearer $MIRAI_API_KEY" \
  -H "Idempotency-Key: call-2026-10-02-0412" \
  -F file=@call.wav \
  -F context="A call to a university admissions desk about a course and its fees" \
  -F vocabulary="Mira University, BBA, Aadhaar"

Questions

Here is what people ask before switching from OpenAI.

What does an hour of audio cost on each?

gpt-transcribe is $0.0045 a minute, which comes to $0.27 an hour, about ₹25.64 at the ECB rate, with no speaker labels. Ours is ₹15 an hour, speaker labels included, so a five-minute call costs ₹1.50.

Does gpt-transcribe support Hindi?

OpenAI's docs print no language list for it. Whisper's list in OpenAI's repository has Hindi, Tamil, Telugu, Gujarati and Marathi, with the note that accuracy varies by language. Whisper itself is removed from the API on February 26, 2027. Test your own recordings before you decide.

What happens to my Whisper integration?

OpenAI retires whisper-1, gpt-4o-transcribe, gpt-4o-mini-transcribe and gpt-4o-transcribe-diarize on February 26, 2027, and points to gpt-transcribe or gpt-live-transcribe. If you have to move anyway, move the Indian calls to us and keep the rest wherever you like.

Is transcription priced like the agents?

No. A whole phone agent on our volume plan is ₹1.50 a minute all-in, with our line and insights. Transcribing a recording costs ₹15 an hour of audio, which is 25 paise a minute.

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: T5 is ₹5 a minute, pay as you go, with our phone line and call insights included. At volume, T1 is ₹1 a minute on ₹1,00,000 a month, and ₹1.50 with our phone line (₹0.30) and call insights (₹0.20). Prices exclude GST.

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

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

  4. Mira Transcribe is ₹15 an hour of recorded audio, charged per second, once, when a job completes; degraded and failed jobs are free. All prices exclude GST.

Get an API key