LecSync vs iFlytek Realtime ASR: CN-EN Captions, Billing and Concurrency

iFlytek's open platform sells realtime speech-to-text as prepaid hour packages on the China product page, with an optional yearly add-on that turns the same WebSocket into live captions in another language. List rates for Chinese-heavy streaming transcription often land under LecSync's dollar-per-minute bill once you buy a mid-size Standard package. The LecSync Realtime Speech API is $0.006 per audio minute ($0.36 per hour) and returns captions, a paired translation across 60 languages, speaker labels and running minutes on one connection — without a separate translation license.

So the choice is less "who wins a pure CNY STT tender" and more "do you need bilingual captions, speakers and notes on a flat international minute rate."

Everything below comes from public pages read on October 16, 2026. For iFlytek: the realtime transcription product page (大模型 and 标准版 price cards, plus the 实时翻译 add-on), the Standard RTASR WebSocket docs, the Realtime ASR LLM docs, and NiuTrans text MT pricing for a DIY glue path. For LecSync: the developers page, pricing and connect docs. CNY→USD columns use an estimated 7.0 yuan per dollar because the China pages list yuan only.

The short answer

For Mandarin (and dialect-heavy) streaming transcription inside iFlytek's China footprint, the open-platform packages are hard to ignore. Standard 套餐四 works out to about ¥1.80 per hour of prepaid audio ($0.26/hr at 7 CNY/USD). The LLM line bundles dialect auto-switch (202 dialects) and stronger speaker tools, though its 1,000-hour card is ¥4,000 ($0.57/hr) — above LecSync for STT alone. Concurrent channels ride with the package (for example 30 on the LLM 1,000-hour card; Standard duration packs default to 50 on the product tip).

For live Chinese↔English captions that also need speakers and minutes, LecSync is the flatter path. iFlytek's Standard API can stream a translation on the same socket after you buy the 实时翻译 yearly entitlement (¥20,000 / year on the public card), and language pairs that do not include Chinese must hop through Chinese. That yearly fee amortizes only when you burn a lot of translated hours. LecSync keeps transcription, any-to-any translation among 60 languages, speakers and digest inside $0.36/hr.

How each session works

LecSync. Your server calls POST https://api.lecsync.com/v1/realtime/connect with the audio format, a language pair and the features you want. It gets back a stream URL and a one-time token. The client pushes 16 kHz mono PCM. Transcript segments, their translations (matched by segment ID), speaker labels with millisecond timestamps (speakers: true) and sections of minutes (digest: true) all come back on that one connection. A network stall is capped at 2 seconds of backlog instead of letting latency climb.

iFlytek Standard RTASR. You open wss://rtasr.xfyun.cn/v1/ws with an HMAC signature, stream 16 kHz PCM in ~40 ms frames, and receive intermediate and final sentences with timestamps. Set transType, transStrategy and targetLang after the console translation entitlement is on. Docs note that English→French style pairs must go English→Chinese→French. roleType=2 turns on role separation (still marked as improving). Hot words and vertical domains (pd=edu, finance, medical, …) are first-class. Hitting the authorized channel count returns error 10800.

iFlytek Realtime ASR LLM. A separate WebSocket (office-api-ast-dx.iflyaisol.com) with lang=autodialect (CN/EN + 202 dialects) or lang=autominor (37 languages, ticket required). role_type=2 enables blind speaker separation; optional voiceprint IDs pin known speakers. A single session is capped at 8 hours of audio. Product marketing puts live translation on the shared RTASR page; the public param path for MT that we cite is the Standard API + 实时翻译 card.

Side-by-side facts (checked October 16, 2026)

What 1,000 hours cost

iFlytek rows use public CNY package unit rates (package price ÷ package hours). USD uses estimated 7.0 CNY/USD. The 实时翻译 row assumes you buy the ¥20,000 yearly card and use about 1,000 translated hours in that year — change the utilization and the add-on per hour moves with it. NiuTrans is only for a DIY STT+text-MT glue path (~50,000 characters of transcript per hour at the ¥48/M list).

The second column is calculated from public cards, not a quote. Business discounts and private deployments can move iFlytek further; those are out of scope here. For a wider vendor grid, see what 1,000 hours cost with translation, speakers and summaries.

Where iFlytek is the better pick

• Chinese STT at prepaid package rates. Mid-to-large Standard packages undercut LecSync's $0.36/hr when you only need the source-language transcript and can commit to a year of hours.
• Dialect coverage inside China. The LLM autodialect mode advertises CN/EN plus 202 Chinese dialects without manual switching — a natural fit for domestic call centers, classrooms and local livestreams.
• Bundled concurrency. Packages publish channel counts (or default to 50 on Standard duration tips). Error codes make over-limit behavior explicit.
• Hot words and vertical domains. Console hotwords and pd tags (edu, finance, medical, court, …) are first-class on both Standard and LLM docs.
• Speaker tools with voiceprint. LLM docs document blind separation plus registered voiceprint IDs when the cast is known.
• Existing 讯飞 stack. If the rest of the product already sits on xfyun.cn auth, SDKs and China endpoints, staying put avoids a second vendor.

Where LecSync fits better

• Captions in a second language across 60 options. Translation streams with the transcript at about +88 ms and comes back paired by segment ID — including pairs that never touch Chinese. More on the trade-off in built-in translation vs DIY STT + MT.
• No yearly translation license. Features on or off do not change $0.006/min; empty connections are not billed.
• Who said what, while it's being said. Every final segment can carry a speaker label with startMs / endMs. See building a meeting client with speakers and timestamps.
• Notes during the meeting. digest: true pushes sections of minutes over the same socket.
• Latency that doesn't drift. After a network stall, backlog is capped at 2 seconds. More in how LecSync caps backlog at 2 seconds.
• Gateways outside a single China endpoint story. Six gateway regions when speakers are not only on the mainland network path.

FAQ

Does iFlytek realtime ASR include translation out of the box?

Transcription packages alone return source-language text. Live translation on the Standard WebSocket needs the console translation entitlement (public card: ¥20,000 per year, 15-day trial) plus transType / targetLang params. FAQ copy focuses on Chinese paired with Japanese, Korean, Russian, French, Spanish, Vietnamese and Cantonese.

Can iFlytek translate English directly into French on one hop?

Standard docs say pairs that do not include Chinese must use Chinese as a bridge (for example English→Chinese→French). LecSync configures one language pair per stream among 60 languages without that bridge rule.

How does concurrency work on iFlytek?

You buy (or receive) a channel count with the package. The LLM 1,000-hour card lists 30 concurrent channels; Standard duration packages tip default 50. Exceeding the grant returns an over-limit error instead of silent queueing.

Does LecSync charge extra for speakers or digest?

No. Realtime transcription in 60 languages, realtime translation between any two of them, audio-aligned timestamps, speaker labels, glossary and domain context, AI translation polish, live structured minutes, and optional 30-day storage all sit inside $0.006 per minute.

Is the global.xfyun.cn page the same price list?

The English product page shows package sizes and concurrent numbers but, on the fetch used for this article, did not expose USD list prices. All $/hr math here comes from the China open-platform CNY cards.

Bottom line

Pick iFlytek when the job is Chinese-heavy streaming transcription (especially dialects) inside the 讯飞 open platform, you can use prepaid hour packages, and live translation is either unnecessary or high enough volume to absorb the yearly 实时翻译 card. Pick LecSync when the live product needs wide-matrix translation, speaker-aware captions and running minutes on one flat dollar-per-minute bill with international gateways. Start with a short proof stream on the developers page — POST /v1/realtime/connect, then one WebSocket.