10 Indian languages + 40+ voices
Hindi (Kavya, Anand, Priya and more), plus Bengali, Gujarati, Kannada, Malayalam, Marathi, Odia, Punjabi, Tamil, Telugu and Indian English. Male and female voices in every language.
Text-to-Speech API
Text-to-speech API with natural Indian-language voices
40+ natural voices across 10 Indian languages plus Indian English. Streaming output, $0.30 per 10K characters. Built for voice agents, audiobooks, IVR, and accessibility.
- 10 Indian languages × 40+ natural voices
- Streaming output for live voice agents
- Speed control, 6 audio formats
- $0.30 / 10K chars, transparent pricing, no seat fees
- Indian languages
- 10
- Natural voices
- 40+
- Real-time output
- Streaming
- per 10K chars
- $0.30
How it works
From text to natural voice in 4 steps
Sample voices, pick one, ship to production.
- 01
Pick a voice
40+ Bulbul voices across 10 Indian languages plus Indian English, male and female. Sample them in-dashboard before deploying.
- 02
Send text
POST your text with a voice and language, or stream chunks. Control speed from 0.25x to 4x.
- 03
Receive audio
MP3, WAV, OPUS, AAC, FLAC, or raw PCM. Batch mode for files; streaming for real-time voice.
- 04
Reuse
The OpenAI-compatible audio.speech shape: one call returns the audio bytes, ready to play or store.
Features
Why teams building Indian products choose CallMissed TTS
Because Google's Hindi voices still sound like a call-center IVR. Ours don't.
Natural Indic prosody
Sarvam's Bulbul voices are tuned for Indian languages, with intonation and rhythm that sound native, not a translated accent.
Streaming output
Stream audio bytes as they're synthesized, fast enough for live voice agents and interactive IVR, or fetch the whole clip in one batch call.
Speed control
Set playback speed from 0.25x to 4x per request. Pick the output format your pipeline needs: MP3, WAV, OPUS, AAC, FLAC, or raw PCM.
OpenAI-compatible shape
Call it with the audio.speech.create pattern from the OpenAI SDK: one endpoint, a voice, a language, your text. No new client to learn.
India data residency
Text and audio are processed in our India region (Azure Central India). DPDP Act-aligned, tenant-isolated per API key.
Use Cases
Text-to-speech use cases
Voice agents, audiobooks, IVR, accessibility, navigation: same API.
Voice AI agents
Natural voice for your phone bot
Pair our TTS with the Voice Agent API for a full voice pipeline. Streaming output means callers hear the reply as it's synthesized, not after an awkward gap.
Result
Voice agents that customers don't hang up on.
Audiobooks + e-learning
Indian-language audiobooks at scale
Convert educational content, books, articles into natural-sounding audio across 10 Indian languages. Students consume lessons while commuting; learners without literacy get access via voice.
Result
Turn a whole library into audio at per-character API rates.
IVR + phone menus
Dynamic IVR prompts without re-recording
Change your IVR greetings, hold messages, or queue announcements in minutes, not weeks waiting for a studio session. Say 'Queue is longer than usual' in Tamil on-demand without re-recording audio files.
Result
Launch campaign greetings in an hour, not a week.
Accessibility
Read-aloud for low-literacy users
Government portals, banking apps, and health apps add a 'Listen' button that reads content in the user's chosen Indian language. Essential for low-literacy populations and visually impaired users.
Result
WCAG AA compliance + genuine accessibility for Bharat.
Navigation & in-app voice
Turn-by-turn directions in local languages
Mapping apps, delivery driver apps, cab aggregators: give turn-by-turn voice prompts in the driver's local language. Works offline when prompts are pre-generated and cached on device.
Result
Voice navigation across 10 Indian languages.
Marketing + announcements
Personalized voice notes at scale
Send hyper-personalized WhatsApp voice notes ('Hi Rajesh, your order is out for delivery') in the customer's language. Higher open rates than text, warmer than generic SMS.
Result
Warmer than SMS, in the customer's own language.
Compare
CallMissed TTS vs Google, AWS, ElevenLabs, Azure
On Indian languages we lead; on price we crush; on API ergonomics we're OpenAI-compatible.
| Feature | CallMissed | Google TTS | AWS Polly | ElevenLabs | Azure TTS |
|---|---|---|---|---|---|
| Native Indian-language voices | |||||
| Natural Indic prosody | |||||
| Streaming output | |||||
| Speed control | |||||
| India data residency | |||||
| OpenAI-compatible shape | |||||
| Pricing per 10K chars | $0.30 | $0.16 | $0.16 | $3.00 | $0.16 |
Comparison based on publicly listed features as of 2026. Check each vendor's site for the latest.
Code
Shipping voice in 5 lines
speak.pyPython
from openai import OpenAI
client = OpenAI(
base_url="https://api.callmissed.com/v1",
api_key="cm_your_key",
)
audio = client.audio.speech.create(
model="bulbul:v3", # Sarvam — 37 voices, 11 Indian languages.
# Also: "aura-2-en" (Deepgram English, 40 voices)
# Also: "aura-2-es" (Deepgram Spanish, 10 voices)
# Also: "melotts" (open-source en + fr, cheapest)
voice="ritu", # Hindi female — or shubh, priya, rahul…
input="नमस्ते, आपका ऑर्डर कल पहुँचेगा।",
response_format="mp3",
)
audio.stream_to_file("greeting.mp3")Python: synthesize Hindi voice, save as MP3
stream.mjsJavaScript
import { CallMissed } from "callmissed";
const cm = new CallMissed({ apiKey: process.env.CM_KEY });
const stream = cm.audio.tts.stream({
model: "bulbul:v3", // or "aura-2-en" / "aura-2-es" / "melotts"
voice: "shubh", // 37 Sarvam speakers; pick a Tamil one for ta-IN
text: "உங்கள் ஆர்டர் நாளை வருகிறது",
format: "opus",
sampleRate: 24000,
});
// pipe to telephony / websocket / audio element
stream.on("data", (chunk) => speaker.write(chunk));Node/JS: stream TTS audio for a live voice agent
FAQ
Text-to-speech API questions, answered
What is a text-to-speech API?
A text-to-speech (TTS) API converts written text into natural-sounding audio. You send a text string, a voice, and a language; it returns audio bytes (MP3/WAV/OPUS and more) you can play in an app, broadcast over the phone, or save as a file. CallMissed's TTS uses Sarvam's Bulbul voices, tuned for Indian languages, so Hindi, Tamil, Marathi etc. sound native, not a translated accent.
Which Indian languages and voices are available?
Sarvam Bulbul covers 10 Indian languages plus Indian English: Hindi, Bengali, Gujarati, Kannada, Malayalam, Marathi, Odia, Punjabi, Tamil, and Telugu. Across them are 40+ named voices, male and female, including Kavya, Anand, Priya, Shubh, Aditya, Ritu and more. Sample every voice in the dashboard before committing.
How natural do the voices sound?
Bulbul voices are tuned specifically for Indian languages, so intonation and rhythm sound native rather than a translated accent. For a live demo, sample any voice in the dashboard and compare it against your current provider on your own script.
Does TTS stream, or only return a full clip?
Both. Fetch the whole clip in one batch call, or stream audio bytes as they're synthesized for live voice agents where a caller speaks, an LLM replies, and TTS must start speaking quickly. Streaming works over the audio endpoint; batch returns the complete file.
How much does text-to-speech cost?
Sarvam Bulbul v3 is $0.30 per 10,000 characters (Bulbul v2 is half that). You pay from your credits (1 credit = ₹1 = $0.01) and only for the characters you synthesize. Free tier includes 1000 API credits to start.
Can I control speed?
Yes. Set the speed parameter between 0.25x and 4x per request. Pitch, SSML markup, and voice cloning are not offered on this API today: pick a voice and language, send text, get audio.
What audio formats does the API return?
MP3, WAV, OPUS, AAC, FLAC, and raw PCM. Choose the format via the response_format parameter, MP3 by default. The batch shape matches the OpenAI audio.speech pattern.
Where is my text processed?
Text and generated audio are processed in our India region (Azure Central India), tenant-isolated per API key and DPDP Act-aligned. We don't train on your content.
How do I integrate it?
Use the OpenAI SDK's audio.speech pattern, or plain REST (curl works). Point base_url at https://api.callmissed.com/v1, send text with a voice and language, and get audio bytes back. Full docs and copy-paste examples at docs.callmissed.com.
Try 40+ Indian voices free
Grab an API key, sample every voice in the dashboard, and ship natural-sounding audio today.