Time to final segment
From 40 ms
Nari Qwen3-ASR 1.7B
Realtime speech recognition with 40 ms time to final segment. Choose Fast for latency or Standard for cost efficiency.
Price per hour
From $0.06
Languages
30 Languages
Choose what matters most
Our Qwen3-ASR endpoint comes in two variants. Use the same WebSocket API and change only the ID.
Standard
For long-running transcription workloads where serving cost matters most.
- Latency
- 300 ms
- Pricing
- $0.06
Fast
For voice agents and interactive applications where every millisecond matters.
- Latency
- 40 ms
- Pricing
- $0.12
Lower latency, lower cost
Based on preliminary independent measurements of publicly available STT endpoints from US East. Results may vary by region, connection, and workload. Pricing checked Sep 8, 2026 from Artificial Analysis.
One model, multilingual speech
Automatic language identification across 30 officially supported languages.
View the official model card30 Supported Languages
- Chinesezh
- Englishen
- Cantoneseyue
- Arabicar
- Germande
- Frenchfr
- Spanishes
- Portuguesept
- Indonesianid
- Italianit
- Koreanko
- Russianru
- Thaith
- Vietnamesevi
- Japaneseja
- Turkishtr
- Hindihi
- Malayms
- Dutchnl
- Swedishsv
- Danishda
- Finnishfi
- Polishpl
- Czechcs
- Filipinofil
- Persianfa
- Greekel
- Hungarianhu
- Macedonianmk
- Romanianro
Audio in. Text out.
Open one WebSocket, stream PCM audio, and receive partial and final transcripts as events.
WebSocket · Python
URL = "wss://api.narilabs.com/v1/realtime?intent=transcription"
async with connect(
URL,
additional_headers={"Authorization": f"Bearer {os.environ['NARI_API_KEY']}"},
) as socket:
await socket.send(json.dumps({
"type": "session.configure",
"session": {"model": "qwen3-asr"},
}))
await socket.recv() # session.configuredBuilt for realtime transcription
Voice agents
Stream partial and final transcripts into realtime conversational systems.
Live captions
Turn ongoing speech into responsive captions for streams, events, and accessibility.
Meetings and call analytics
Create searchable transcripts for notes, review, and downstream analysis.
Bring realtime transcription to your product
Share your audio, language requirements, and latency targets with our engineers.