# Frequency by Infinite Inference

> Voice AI built and run in India. Agents from ₹1/min with a volume commitment. Try the sandbox.

Frequency brings ASR, LLM, TTS and orchestration together for Indian voice conversations.
Fourier TTS is live in production. Hindi, English, Gujarati, Tamil and Telugu are live in speech recognition and text to speech, and so is Hinglish. Bengali, Marathi and Kannada are next.
Every company starts on T5: ₹5/minute, no monthly commitment. T3: ₹3/minute with a ₹50,000 monthly minimum. T1: ₹1/minute with a ₹1,00,000 monthly minimum, optional Mirai telco at ₹0.30/minute and optional insights at ₹0.20/minute. The same models and features, including single-prompt and node-based agents, are available on every tier. T5 and T3 include Mirai phone calling and insights. T1 adds ₹0.30/minute only for Mirai-managed phone calls. BYO telco, WhatsApp and browser calls have no Mirai telco charge on any tier. Provider charges you pay directly are separate. All prices exclude GST.
Text to speech: ₹1.80 per 1,000 characters. Monthly commitments are credited to your wallet. Unused credit never expires.
LLM: ₹15 input, ₹9.30 cached input and ₹55 output per 10,00,000 tokens. Pay as you go: 1,000 requests/minute, no monthly commitment. Business: 5,000 requests/minute with a ₹50,000 monthly commitment. Rate limits apply per workspace across all API keys. Monthly commitments are credited to your wallet. Unused credit never expires. All prices exclude GST.

For current API details, consult the documentation. More information about the models is coming soon.

## Website

- [Home](https://scalefrequency.com/): the voice stack, browser demo, production story and FAQ.
- [Pricing](https://scalefrequency.com/pricing): API rates, monthly commitments, telco and insights add-ons, and billing examples.
- [AI calling agents](https://scalefrequency.com/ai-calling): AI agents that make and take calls in Hindi, English, Gujarati, Tamil and Telugu, from ₹1 a minute. Pay as you go at ₹5, no monthly minimum.
- [Speech APIs for developers](https://scalefrequency.com/developers): Fourier TTS, STT and LLM APIs for Hindi, English, Gujarati, Tamil and Telugu. One key, a five-minute quickstart and ₹200 of sandbox credit.
- [Compare](https://scalefrequency.com/compare): Frequency beside Vapi, Retell AI, Bland AI, Bolna, Gnani.ai and Skit.ai, each cell from that company's own public pages, checked 8 and 17 October 2026.
- [Fourier TTS](https://scalefrequency.com/models/text-to-speech): Hindi text-to-speech; Voice Arena Hindi Elo 993 vs Sarvam Bulbul V3 959.
- [Fourier STT](https://scalefrequency.com/models/speech-to-text): Hindi speech-to-text; 9.12% normalised WER on the 10,000-utterance RinggAI Hindi benchmark.
- [Use cases](https://scalefrequency.com/use-cases): sample calls, upload columns, results and cost per call, by industry.
- [Alternatives](https://scalefrequency.com/alternatives): Frequency beside one platform at a time, their column from their own pages.
- [Tools](https://scalefrequency.com/tools): cost calculators for AI calling and text to speech.
- [Languages](https://scalefrequency.com/languages): what is live, what is next, and a page per live language with real voice samples.
- [Glossary](https://scalefrequency.com/glossary): voice AI and Indian calling terms, each in plain words with an example call.
- [Who we are](https://scalefrequency.com/about): why we build and run voice infrastructure in India.
- [Developer guides](https://scalefrequency.com/guides): Hindi TTS, speech to text, Realtime API, phone calls and campaigns from code, provider comparisons, and Gemma/vLLM serving measurements.
- [Journal](https://scalefrequency.com/journal): engineering decisions and measurements.
- [Privacy](https://scalefrequency.com/privacy): optional analytics, demo audio and bookings.
- [Terms](https://scalefrequency.com/terms): the terms for the sandbox, console and API, including how call data is processed and kept.

## Developer guides

- [Making Gemma 4 26B A4B fast enough for a voice conversation](https://scalefrequency.com/guides/gemma-4-26b-a4b-low-latency): A Gemma voice-serving case study: 1,520 measured requests, concurrency trade-offs, connection reuse and a downloadable analysis kit.
- [How FP4 can help voice inference—and what it cannot promise](https://scalefrequency.com/guides/fp4-quantization-voice-inference): Understand NVFP4 with Gemma 4 26B A4B: weight memory, hardware kernels, FP8 KV cache and the experiment needed to prove a latency improvement.
- [Tune vLLM for voice latency, not just tokens per second](https://scalefrequency.com/guides/vllm-throughput-voice-agents): Use our archived concurrency sweep to understand vLLM batching, tail latency, prefill budgets, warmup and goodput for voice workloads.
- [Manage voice-agent context before buying more inference capacity](https://scalefrequency.com/guides/voice-agent-context-management): Measured shared-prefix results, KV capacity arithmetic and a practical context budget for voice agents using Gemma and vLLM.
- [Prompt-based or node-based voice agent: how to choose](https://scalefrequency.com/guides/prompt-based-vs-node-based-voice-agents): One prompt or a graph of nodes? Choose by branching, risk and recovery: how our node-based agents run, an appointment example and test cases.
- [How to fix an overloaded voice-agent prompt](https://scalefrequency.com/guides/fix-overloaded-voice-agent-prompts): Fix contradicting instructions, move state and permissions into code, give each turn only what it needs, read the agent's logs and test corrections.
- [Voice AI APIs for developers building in India](https://scalefrequency.com/guides/voice-ai-api-india): Which Frequency API to start with: phone agents, the Realtime API, Fourier TTS or transcription. What each one does, and how a call works.
- [Integrated voice API or build your own stack?](https://scalefrequency.com/guides/voice-ai-stack-build-vs-buy): A hosted voice-agent API, a speech-to-speech Realtime API, or a stack you build from ASR, LLM and TTS: control, latency, cost and who runs it.
- [Choosing a Hindi and Hinglish text-to-speech API](https://scalefrequency.com/guides/hindi-text-to-speech-api): How to judge Fourier TTS on Hindi and Hinglish: script, code-switching, numbers, the four voices, streaming for phone lines and a starter test set.
- [Use Fourier TTS with an OpenAI-compatible speech client](https://scalefrequency.com/guides/openai-compatible-tts): On OpenAI's speech API? Change the base URL, model and voice, and your code speaks Hindi, Tamil, Telugu and more with Fourier TTS. Formats and streaming.
- [Connect voice-agent webhooks to your application](https://scalefrequency.com/guides/voice-agent-webhooks): Hear about every call the moment it changes. Verify the signature on the raw body, store before you answer, ignore retried duplicates, order by status.
- [How to measure latency in a voice AI application](https://scalefrequency.com/guides/voice-ai-latency): Split the delay a caller hears into turn detection, the model, TTS first byte, the network and playback, and read what Mira's request log reports.
- [Voice AI API pricing: minutes, characters and commitments](https://scalefrequency.com/guides/voice-ai-api-pricing): How we bill voice agents, Realtime sessions, TTS and transcription: rates, 10-second blocks, what's free, commitments, and worked monthly estimates.
- [Make an AI phone call from Python with the Mirai API](https://scalefrequency.com/guides/ai-phone-call-api-python): Place a real AI phone call in Hindi, Tamil or Telugu from Python, then read the transcript off a signed webhook. Three requests, with httpx and Flask.
- [Run bulk AI calls from a CSV with a Mirai campaign](https://scalefrequency.com/guides/bulk-ai-calls-from-csv): Upload a CSV this afternoon and an AI agent calls every row in Hindi, Tamil, Telugu and more, inside your hours, with retries and a result per contact.
- [Hindi and Hinglish TTS in Pipecat with pipecat-mirai](https://scalefrequency.com/guides/pipecat-hindi-tts): Swap one service and your Pipecat bot speaks Hindi and Hinglish. pipecat-mirai 0.5.1: install, pipeline, an 8 kHz phone setup, Realtime and fixes.
- [Hindi speech-to-text API: transcribe calls with Mira](https://scalefrequency.com/guides/hindi-speech-to-text-api): Send a Hindi or Hinglish call recording and get back who said what, with timestamps. The /v2/transcriptions request, polling, speakers and ₹18 an hour.
- [Hindi text-to-speech APIs compared: prices and features, October 2026](https://scalefrequency.com/guides/hindi-tts-api-comparison): Published prices and Hindi support for Fourier TTS, Sarvam Bulbul, ElevenLabs, Google, Azure, Amazon Polly, OpenAI and Cartesia, each sourced and dated.
- [Hindi speech-to-text APIs compared: prices and features, October 2026](https://scalefrequency.com/guides/hindi-speech-to-text-api-comparison): Published prices and Hindi support for Frequency, Sarvam, ElevenLabs, Deepgram, Google, Azure, Amazon Transcribe and OpenAI, sourced and dated.
- [Realtime speech-to-speech API for Hindi voice agents](https://scalefrequency.com/guides/realtime-speech-to-speech-api): Send a caller's audio over one WebSocket and get the reply spoken in Hindi, English, Gujarati, Tamil or Telugu. OpenAI Realtime protocol, barge-in, tools.
- [Voice AI infrastructure in India: TTS, STT, LLM and agent APIs](https://scalefrequency.com/guides/voice-ai-providers-india): Who sells TTS, STT, LLM and agent APIs in India: published prices, Hindi support and where the audio is processed, read on 16 October 2026.

## Use cases

- [Confirm every COD order before it ships](https://scalefrequency.com/use-cases/cod-order-confirmation): An AI agent calls every cash-on-delivery order in the customer's own language, confirms it and fixes the address before you pack. From ₹1 a minute.
- [Win back the carts that stopped at a question](https://scalefrequency.com/use-cases/abandoned-cart-recovery): An AI agent calls shoppers who left a cart, answers the size, COD or delivery doubt that stopped them, and offers the checkout link. From ₹1 a minute.
- [Save failed deliveries before they become returns](https://scalefrequency.com/use-cases/ndr-delivery-failure-calls): When a courier marks a parcel undelivered, an AI agent calls for a new day, a landmark or a cancellation, and flags disputed attempts. From ₹1 a minute.
- [Call every new lead while it's still warm](https://scalefrequency.com/use-cases/lead-qualification): An AI agent calls each new form or ad lead, often within minutes, asks your qualifying questions and books the site visit or demo. From ₹1 a minute.
- [Know who will pay on time, days before the due date](https://scalefrequency.com/use-cases/payment-reminders): A polite AI call a few days before an EMI or invoice is due. It confirms the amount, notes the promised date and flags disputes. From ₹1 a minute.
- [Know by evening who's coming in tomorrow](https://scalefrequency.com/use-cases/appointment-reminders): An AI agent calls tomorrow's patients, confirms each time, and moves the booking in your calendar when someone can't come. From ₹1 a minute.
- [Turn the monthly due list into booked service slots](https://scalefrequency.com/use-cases/service-due-reminders): An AI agent calls owners whose vehicle or appliance service is due, books a slot at their usual centre, and notes why when they say no. From ₹1 a minute.
- [Find out why a customer gave you a 3](https://scalefrequency.com/use-cases/customer-feedback-surveys): A day after a delivery or service visit, an AI agent asks for a score and the reason, and flags anything still broken. From ₹1 a minute.
- [Answer the phone when nobody's at the desk](https://scalefrequency.com/use-cases/inbound-receptionist): An AI agent answers your business number, handles the questions you hear every day, and takes a proper message for your team. From ₹1 a minute.
- [Catch renewals before the policy lapses](https://scalefrequency.com/use-cases/renewal-reminders): An AI agent calls before a policy or plan expires, gives the date and amount, offers the renewal link and books an advisor. From ₹1 a minute.
- [Screen hundreds of applicants without a day on the phone](https://scalefrequency.com/use-cases/recruitment-screening): An AI agent calls every applicant with the same questions: area, joining date, experience, pay. Recruiters get a sheet to sort. From ₹1 a minute.
- [Call back every enquiry before another institute does](https://scalefrequency.com/use-cases/admissions-enquiry-follow-up): An AI agent calls back every course enquiry, answers fee and batch questions for the student or the parent, and books a counsellor. From ₹1 a minute.

## Alternatives

- [Looking for a Sarvam alternative?](https://scalefrequency.com/alternatives/sarvam): Sarvam vs Frequency: Fourier TTS at ₹1.80 per 1,000 characters, speech to text, and phone agents at ₹1 a minute. Both in rupees, side by side.
- [Looking for an ElevenLabs alternative for India?](https://scalefrequency.com/alternatives/elevenlabs): ElevenLabs speaks 90+ languages, billed in dollars. We speak Hindi, Gujarati, Tamil, Telugu and English, in rupees: agents from ₹1 a minute.
- [Looking for a Vapi alternative for Indian calls?](https://scalefrequency.com/alternatives/vapi): On Vapi, a minute is five prices added up, in dollars. Our Indian-language phone agents have one price, in rupees: ₹1 a minute. See both side by side.
- [Looking for a Bolna alternative?](https://scalefrequency.com/alternatives/bolna): Bolna and Frequency both build phone agents for India. Ours include the line on pay as you go, bill in 10-second blocks, and drop to ₹1 a minute.
- [Looking for a Retell AI alternative for India?](https://scalefrequency.com/alternatives/retell-ai): Retell AI's own numbers are US and Canada only. Ours are Indian, the line is included on pay as you go, and our agents cost ₹1 a minute.
- [Looking for a Bland AI alternative?](https://scalefrequency.com/alternatives/bland-ai): Bland AI and Frequency both sell the call at one rate. Ours is in rupees, includes the line on pay as you go, and is ₹1 a minute on our volume plan.
- [Looking for a Synthflow alternative with a price you can read?](https://scalefrequency.com/alternatives/synthflow): Synthflow sells enterprise contracts from $30,000 a year, rate inside the contract. Our Hindi phone agents have a published price: ₹1 a minute.
- [Looking for a LiveKit alternative that comes with the agent?](https://scalefrequency.com/alternatives/livekit): LiveKit is infrastructure you build an agent on; models and phone line are billed on top. Ours is the whole call, one price in rupees: ₹1 a minute.
- [Looking for a Twilio ConversationRelay alternative for India?](https://scalefrequency.com/alternatives/twilio-conversationrelay): On Twilio, a call to India is ConversationRelay, the call leg and your own model, in dollars. Our Hindi phone agents cost one price in rupees: ₹1 a minute.
- [Looking for an OpenAI Realtime API alternative for Indian calls?](https://scalefrequency.com/alternatives/openai-realtime): OpenAI prices speech-to-speech per audio token, in dollars, with no Indian-language list. Our Hindi phone agents cost ₹1 a minute, line included.
- [Looking for a Dialogflow CX alternative for Indian calls?](https://scalefrequency.com/alternatives/dialogflow-cx): Dialogflow CX bills voice agents per second of audio, in dollars, with US-only numbers. Our Hindi phone agents cost ₹1 a minute, line included.
- [Looking for an Amazon Connect alternative for Indian calls?](https://scalefrequency.com/alternatives/amazon-connect): Amazon Connect is three AWS products billed apart, and India is off its default calling lists. Our Hindi phone agents cost ₹1 a minute, line included.
- [Looking for a Gnani.ai alternative with a price on the page?](https://scalefrequency.com/alternatives/gnani): Gnani.ai sells Indian-language voice agents through a demo call. We print the price: AI agents at ₹1 a minute, a sandbox with no sign-up, live today.
- [Looking for an OpenAI TTS alternative with Indian voices?](https://scalefrequency.com/alternatives/openai-tts): Keep your OpenAI code and change the base URL. Fourier TTS takes the same /v1/audio/speech request, with Indian voices at ₹1.80 per 1,000 characters.
- [Looking for a Google Cloud TTS alternative for Hinglish?](https://scalefrequency.com/alternatives/google-cloud-tts): Google has 30 Hindi Chirp 3: HD voices and a free tier. Ours are voices built for Hinglish calls, made in India, at ₹1.80 per 1,000 characters.
- [Looking for an Azure Speech alternative for Indian calls?](https://scalefrequency.com/alternatives/azure-speech): Azure covers many Indian languages from Central India. We focus on the phone call: voices at ₹1.80 per 1,000 characters, call transcripts at ₹18 an hour.
- [Looking for an Amazon Polly alternative for Hindi?](https://scalefrequency.com/alternatives/aws-polly): Polly has two Hindi voices, Aditi and Kajal. Fourier TTS has four, made for phone calls in India, at ₹1.80 per 1,000 characters.
- [Looking for a Deepgram alternative for Indian calls?](https://scalefrequency.com/alternatives/deepgram): Transcribe recorded calls at ₹18 an hour, speaker labels included, with an Indian voice on the same key. Deepgram and us, side by side, in rupees.
- [Looking for a Cartesia alternative, billed in rupees?](https://scalefrequency.com/alternatives/cartesia): Cartesia sells Sonic on monthly credit plans in dollars. Fourier TTS is pay per character in rupees, ₹1.80 per 1,000, with voices built for Indian calls.
- [Looking for a Smallest.ai alternative?](https://scalefrequency.com/alternatives/smallest-ai): Smallest.ai and Frequency side by side, in rupees: text to speech per 10 lakh characters, speech to text per hour, and phone agents at ₹1 a minute.
- [Looking for a Murf alternative for phone calls?](https://scalefrequency.com/alternatives/murf): Murf's Falcon is $0.01 per 1,000 characters, below Fourier TTS at ₹1.80. What the gap buys: voices for Indian calls, GPUs in India, agents on one key.
- [Looking for an AssemblyAI alternative for Indian calls?](https://scalefrequency.com/alternatives/assemblyai): AssemblyAI prices by the hour in dollars, speaker labels extra. We transcribe Indian calls at ₹18 an hour, speaker labels included, on our GPUs in India.
- [Looking for a Speechmatics alternative for Indian calls?](https://scalefrequency.com/alternatives/speechmatics): Speechmatics' lowest rate needs a training opt-in and a yearly commitment. We transcribe Indian calls at ₹18 an hour, speaker labels in, on GPUs in India.
- [Looking for a Whisper or gpt-transcribe alternative for Hindi calls?](https://scalefrequency.com/alternatives/openai-stt): Whisper retires in February 2027; gpt-transcribe lists no languages. We transcribe Indian calls at ₹18 an hour, speaker labels included, on GPUs in India.

## Languages

- [Hindi](https://scalefrequency.com/languages/hindi): Hindi phone agents, Hindi text to speech and Hindi transcription, on our own GPUs in India. Hear four voices, then try a call in the sandbox.
- [Hinglish](https://scalefrequency.com/languages/hinglish): Hinglish phone agents and text to speech that switch between Hindi and English mid-sentence, like your customers do. Hear four voices.
- [Indian English](https://scalefrequency.com/languages/english): Indian English phone agents and text to speech, in the same four voices as our Hindi. Hear each one, then try a call in the sandbox.
- [Gujarati](https://scalefrequency.com/languages/gujarati): Gujarati phone agents, Gujarati text to speech and Gujarati transcripts, on our own GPUs in India. Hear four voices, then try a call.
- [Tamil](https://scalefrequency.com/languages/tamil): Tamil phone agents, Tamil text to speech and Tamil transcripts, on our own GPUs in India. Hear four voices, then try a call in the sandbox.
- [Telugu](https://scalefrequency.com/languages/telugu): Telugu phone agents, Telugu text to speech and Telugu transcripts, on our own GPUs in India. Hear four voices, then try a call in the sandbox.

## Tools

- [AI calling cost calculator](https://scalefrequency.com/tools/ai-calling-cost-calculator): What your AI calls would cost a month, from ₹1 a minute or ₹5 pay as you go, and which plan is cheapest for you.
- [Text-to-speech cost calculator](https://scalefrequency.com/tools/tts-cost-calculator): What Fourier TTS would cost for your text, in characters or words, at ₹1.80 per 1,000 characters, and how many requests each plan runs at once.

## Glossary

- [10-second billing blocks (pulse)](https://scalefrequency.com/glossary/10-second-billing-blocks): A billing block, or pulse, is the unit a call's length is rounded up to before it's charged. With 10-second blocks, a call is billed in small slices instead of whole minutes.
- [ASR vs STT](https://scalefrequency.com/glossary/asr-vs-stt): ASR (automatic speech recognition) and STT (speech to text) are two names for the same job: turning spoken audio into written words. Researchers tend to say ASR, and API products tend to say STT.
- [Backchannel](https://scalefrequency.com/glossary/backchannel): A backchannel is a short sound a listener makes to show they're following, like "हाँ", "जी" or "hmm". It doesn't ask for the turn, so the speaker carries on.
- [Barge-in](https://scalefrequency.com/glossary/barge-in): Barge-in is when the caller starts talking while the voice agent is still speaking, and the agent stops to listen. Without it, people have to sit through every sentence before they can answer.
- [BYO telephony (BYOC)](https://scalefrequency.com/glossary/byo-telephony): BYO telephony, or bring your own carrier (BYOC), means the voice AI platform makes and takes calls on your phone numbers and your telecom account. You keep the numbers, the carrier contract and the carrier's bill, and the platform supplies the agent.
- [Caller ID (CLI) and number pools](https://scalefrequency.com/glossary/caller-id): Caller ID, or CLI (calling line identification), is the number a person sees when your call rings. A number pool spreads your outbound calls across several numbers, so no single one calls too many people too fast.
- [Caller-perceived latency](https://scalefrequency.com/glossary/caller-perceived-latency): Caller-perceived latency is the silence a caller hears between finishing their sentence and hearing the agent reply. It's the delay that counts, because it's the only one the caller experiences.
- [Character error rate (CER)](https://scalefrequency.com/glossary/character-error-rate): Character error rate, or CER, scores a speech recogniser letter by letter instead of word by word. A small spelling slip costs a little, where word error rate would count the whole word as wrong.
- [Code-switching (Hinglish)](https://scalefrequency.com/glossary/code-switching): Code-switching is moving between two languages inside one conversation, often inside one sentence. Hinglish, the mix of Hindi and English you hear on so many Indian business calls, is the everyday example.
- [Concurrency (calls and streams)](https://scalefrequency.com/glossary/concurrency): Concurrency is how many calls, or speech streams, a system runs at the same moment. It's the number of lines open at once, and with call length it sets how many calls you can finish in an hour.
- [DND and the NCPR (India)](https://scalefrequency.com/glossary/dnd): DND, short for Do Not Disturb, is the preference a mobile user in India registers with their telecom operator to block promotional calls and messages. It began as the National Customer Preference Register (NCPR); under TRAI's current rules the preferences sit on the operators' shared DLT platform.
- [Elo rating (Voice Arena)](https://scalefrequency.com/glossary/elo-rating): An Elo rating ranks voices by blind head-to-head votes. Listeners hear two clips of the same text without knowing who made them, pick the better one, and every vote nudges the two ratings apart.
- [End-of-turn detection](https://scalefrequency.com/glossary/end-of-turn-detection): End-of-turn detection predicts whether the caller has finished what they wanted to say and is handing the conversation back. It listens to the words and the voice as well as the silence.
- [Endpointing](https://scalefrequency.com/glossary/endpointing): Endpointing is how a speech system decides that the caller has stopped speaking. It is the moment the agent starts working on its reply.
- [μ-law and A-law (8 kHz phone audio)](https://scalefrequency.com/glossary/mu-law-a-law): μ-law and A-law are the two standard ways phone networks squeeze voice into one byte per sample, defined in ITU-T G.711. A phone call carries 8,000 of those samples every second.
- [Node-based agent](https://scalefrequency.com/glossary/node-based-agent): A node-based agent, or flow, runs a call as a map of steps. Each step has its own instructions, and the model moves to the next step when a condition in plain language is met.
- [Post-call analysis](https://scalefrequency.com/glossary/post-call-analysis): Post-call analysis reads a call after it ends and turns the conversation into something you can act on: a short summary, a yes or no against your goal, and the fields you asked for.
- [Prompt-based agent](https://scalefrequency.com/glossary/prompt-based-agent): A prompt-based agent runs the whole call from one written brief: who it is, what it knows, and what it should get done. The language model decides each reply from that brief and the conversation so far.
- [SIP trunk](https://scalefrequency.com/glossary/sip-trunk): A SIP trunk is a phone line delivered over the internet: a connection from a carrier that lets software place and receive ordinary phone calls. One trunk can carry many calls at once.
- [Speech-to-speech and realtime APIs](https://scalefrequency.com/glossary/speech-to-speech-api): A speech-to-speech API takes the caller's voice in and sends the agent's spoken reply back, on one live connection. You stream audio in and play audio out; listening, thinking and speaking all happen on the provider's side.
- [SSML](https://scalefrequency.com/glossary/ssml): SSML, the Speech Synthesis Markup Language, is a W3C standard for adding instructions to text before a voice reads it. Tags can ask for a pause, stress a word, or say how a number or date should be read.
- [Streaming TTS](https://scalefrequency.com/glossary/streaming-tts): Streaming text-to-speech sends audio back in small pieces while it's still being made, so playback can start before the whole sentence is ready. Live voice agents depend on it.
- [TCCCPR and A2P voice calls](https://scalefrequency.com/glossary/tcccpr-a2p-calls): TCCCPR is TRAI's rulebook for commercial calls and messages in India. An A2P (application-to-person) call is one that software places without a person dialling it, and TRAI's 2026 amendment gives it rules of its own.
- [Text normalisation](https://scalefrequency.com/glossary/text-normalisation): Text normalisation turns written shorthand into the words a voice should say. Digits, dates, currency amounts, abbreviations and codes all need it before text-to-speech can read them out.
- [Time to first byte (TTS)](https://scalefrequency.com/glossary/time-to-first-byte): Time to first byte, or TTFB, is how long a text-to-speech service takes to send back the first piece of audio after you ask for it. On a live call it decides how soon the caller hears the agent start speaking.
- [Tool calling in voice agents](https://scalefrequency.com/glossary/tool-calling): Tool calling lets a voice agent reach your systems during a conversation, such as checking an order or booking a slot. The model decides when it needs a tool, asks for it with the right inputs, and uses the answer in its reply.
- [Turn-taking](https://scalefrequency.com/glossary/turn-taking): Turn-taking is how people in a conversation share the floor: one speaks, then the other, with short gaps and little overlap. A voice agent has to do it as well as a person, or the call feels off within seconds.
- [Voice activity detection (VAD)](https://scalefrequency.com/glossary/voice-activity-detection): Voice activity detection, or VAD, is the part of a voice system that tells speech apart from silence and background noise, many times a second. Everything else in the call waits on its answer.
- [Webhook (call events)](https://scalefrequency.com/glossary/webhook): A webhook is a message the calling platform sends to your server the moment something happens on a call: it started, it ended, it went to voicemail, or its analysis is ready. Your system reacts without having to keep asking.
- [Word error rate (WER)](https://scalefrequency.com/glossary/word-error-rate): Word error rate, or WER, measures how often a speech recogniser gets words wrong. It counts the words it swapped, missed or added, and divides by the number of words that were really said.

## Build with us

- [Documentation](https://docs.miraiminds.co): API reference and quickstarts.
- [Fourier TTS quickstart](https://docs.miraiminds.co/v2/tts-quickstart): speech synthesis API.
- [Sandbox](https://sandbox.voice.miraiminds.co): build and test a voice agent.
- [Book a call](https://cal.com/sneh-mehta/15min): discuss your use case with Sneh Mehta.

## Optional

- [Published product overview](https://scalefrequency.com/llms-full.txt): product facts and FAQ in plain text.
- [RSS](https://scalefrequency.com/rss.xml): journal updates.

Contact: sneh@miraiminds.co