科大讯飞开放平台 API

iFlytek Open Platform API provides China's leading speech and AI services — speech recognition (ASR), text-to-speech (TTS), Spark LLM, OCR, translation, voiceprint.

Visit site ↗Documentation ↗

Use it when

#1 industry accuracy in Chinese speech recognition

Watch for

English / other foreign language support weaker than dedicated foreign-language ASR

First check

Register at xfyun.cn to create app for APPID + APIKey + APISecret. Speech recognition via WebSocket: wss://iat-api.xfyun.cn/v2/iat

Auth

api_key

CORS

HTTPS

Yes

Signup

Protocol

REST, WebSocket

Pricing

freemium

Uptime · 30-day window

Health history will appear here after the next daily check.

About this API

iFlytek is an AI company founded 1999 in Hefei, Anhui, China, focused on Chinese speech technology — China's first-tier AI (with Baidu, Alibaba, Tencent, Huawei). Differentiator is "best at Chinese" — years of technical accumulation in Chinese ASR, TTS, dialect recognition makes it nearly monopolize the professional Chinese market. Product coverage: speech recognition (ASR, call center transcription most common), text-to-speech (TTS, smart speakers + audiobooks), Spark LLM (Chinese LLM benchmarking GPT-4), OCR, translation, voiceprint, oral assessment (core tech for education products). Customers include car infotainment (Li Auto, NIO use it), smart speakers (some Tmall Genie tech), education products (multiple electronic dictionaries / learning machines in China), enterprise customer service. Chinese developers building Chinese voice apps default to iFlytek; overseas / non-Chinese scenarios use OpenAI Whisper / AssemblyAI.

What you can build

1Chinese ASR (call center transcription, meeting notes)
2Chinese TTS (smart speakers, audiobook production)
3In-car voice interaction (iFlytek is mainstream auto supplier)
4Education products (oral assessment, homework grading)

Strengths & limitations

Strengths

#1 industry accuracy in Chinese speech recognition
Broad dialect support (Cantonese, Sichuanese, Northeastern, etc.)
On-prem deployment options (compliance-sensitive scenarios)

Limitations

English / other foreign language support weaker than dedicated foreign-language ASR
High latency for overseas business (servers mainly in China)
Complex pricing structure (multi-dimensional by service type / call count / duration)

Example request

Generic template — replace <endpoint> with the real path from the docs.

curl https://www.xfyun.cn/<endpoint> \
  -H "Authorization: Bearer $API_KEY"
# Some providers use X-Api-Key instead — verify in the docs.

Getting started

Register at xfyun.cn to create app for APPID + APIKey + APISecret. Speech recognition via WebSocket: wss://iat-api.xfyun.cn/v2/iat

FAQ

Is English ASR also accurate?+

English isn't iFlytek's main battlefield. Chinese scenarios definitely use iFlytek; English / multilingual scenarios prefer OpenAI Whisper / AssemblyAI.

Which dialects are supported?+

Cantonese, Sichuanese, Northeastern, Henan, Hokkien, Shanghainese, and other mainstream dialects with accurate recognition. Ethnic minority languages (Tibetan, Uyghur) supported but less accurate than Mandarin.

Technical details

CORS: ?HTTPS: YesSignup: ?Open source: No

Auth type: api_key
Pricing: freemium
Rate limit: 免费每日 quota；商业按调用次数计费
Protocols: REST, WebSocket
SDKs: java, python, javascript, csharp, ios, android, mini-program