Chariot-1-ASR Realtime
Stream the audio of a class, an event or a broadcast and get captions while people are still talking, in Hindi, English and 10 more Indian languages, English words and all.
Free to start · No card · WebSocket API · 12 languages
The clip
English words
Clips: FLEURS, CC BY 4.0
#1in all 10 Indian languages, against every streaming API we tested
17%fewer errors than the next best streaming API, without being told the language
3 ha streaming session, with text sent as each piece is recognised
Autolanguage detection, when you do not know the speaker’s language in advance
Benchmark · October 2026
Word error rate by language, lower is better. Chariot was never told the language. Next best is the best of the other streaming APIs in that language.
| Language | Chariot | Next best | Fewer errors |
|---|---|---|---|
| Hindi | 12.90 | 16.76 | 23% |
| Marathi | 19.86 | 20.99 | 5% |
| Bengali | 16.43 | 22.21 | 26% |
| Tamil | 24.09 | 30.13 | 20% |
| Telugu | 20.80 | 22.86 | 9% |
| Kannada | 24.21 | 29.86 | 19% |
| Malayalam | 31.24 | 35.70 | 12% |
| Gujarati | 22.35 | 26.59 | 16% |
| Punjabi | 20.49 | 25.03 | 18% |
| Odia | 21.97 | 28.02 | 22% |
How it works
Connect to the WebSocket in balanced mode, the one that sends text sooner. Set the language, or leave it out.
Stream it in 100 ms chunks from the microphone, mixer or broadcast feed, as 16-bit PCM from 8 to 48 kHz.
Each transcript.delta event carries the newest words. Add them to your player, overlay or page as they arrive.
API
One WebSocket. Send audio as you capture it, read text as it is recognised, and send input.done for the complete transcript.
# pip install websockets
import json, websockets
url = ("wss://api.chariot.in/v1/asr/ws"
"?audio_format=pcm_16000&mode=balanced")
headers = {"chariotai-api-key": API_KEY}
async with websockets.connect(url, additional_headers=headers) as ws:
await ws.recv() # session.started
for chunk in microphone(): # 100 ms of audio at a time
await ws.send(chunk)
await ws.send(json.dumps({"type": "input.done"}))
async for message in ws:
event = json.loads(message)
if event["type"] == "transcript.delta":
print(event["text"], end="")Built for live audio
Balanced sends text sooner, for captions people read as they listen. High accuracy waits for more context.
Codemix keeps borrowed English words in Latin letters; native writes them in the script of the spoken language.
A session runs for up to three hours. For a longer event, open a new session before it ends.
16-bit PCM from 8 to 48 kHz, or 8 kHz mu-law and A-law from a phone line.
Each session is saved to your workspace with its transcript and cost.
You pay for the audio you send, not for the time the connection is open.
Pricing
Speech to text uses the same credits as text to speech, at one rate in both modes. Each plan covers this much audio a month.
| Plan | A month | Speech to text |
|---|---|---|
| Free10,000 credits | ₹0$0 | about 2 hours |
| Starter100,000 credits | ₹499$5 | about 19 hours |
| Creator250,000 credits | ₹999 | about 46 hours |
| Startup1,500,000 credits | ₹4,999$49 | about 280 hours |
| Scale10,000,000 credits | ₹29,999$299 | about 1,900 hours |
Monthly prices. Prices exclude applicable taxes.
Use cases
Voice agents, too: they hear a caller while they are still speaking, in their language.
Upload a recording or a video and get its transcript, in the language it was spoken.
Audio and video to text ›Support and sales calls written down, live from the phone line or from recordings.
Call transcription ›Speak into the microphone, watch the words appear, and copy them anywhere.
Voice typing ›Your data
Speech to text runs on Indian infrastructure. Audio you send and its transcripts are processed in India under the Digital Personal Data Protection Act. Chariot is one of four teams selected under the IndiaAI Mission to build the country’s foundation models.
Text arrives while people are still talking. Balanced mode sends it sooner; high accuracy waits for more context first. How soon it reaches you also depends on your network and how you send the audio.
Yes, if you can send its audio to the API. Chariot returns the text over the WebSocket, and your player, overlay or page shows it.
English words in Indian-language speech are normal input. Choose codemix or native for how they are written. If you do not know the language in advance, leave it out and the model detects it from the audio.
Up to three hours. Open a new session for a longer event.
1.5 credits per second of audio, 90 credits a minute, from the same credits as text to speech. Every account starts with 10,000 free credits, almost two hours of audio.
In India. Audio you send and its transcripts are processed on Indian infrastructure, under the Digital Personal Data Protection Act.
Hindi, Marathi, Bengali, Tamil, Telugu, Kannada, Malayalam, Gujarati, Punjabi, Odia, Sanskrit and English. Set the language for a session, or leave it out and the model detects it from the audio.
Start free with 10,000 credits, almost two hours of audio. No credit card.
