Product · Pipelines

The Five Pipelines

ASR → MT → TTS Pick once, switch instantly

Each pipeline is a curated combination of automatic speech recognition (ASR), machine translation (MT) and text-to-speech (TTS). You select a pipeline via the chip in the app header; SpeechKey handles authentication and routing. All pipelines are optimised for ZH↔EN/DE.

00Overview

Pipeline Latency VPN Cost/min
Standard~1.5 sRequired~$0.02
Pro Live~2.5 sRequired~$0.10
Realtime< 1 sRequired~$0.15
China-Only~2 sNot needed~$0.01
Watch-Modus~1.5 sRequired~$0.02

01Standard

The default for most situations. Reliable in long business meetings, solid quality, low cost. If you don't know which pipeline to choose: use Standard.

ASR
iFLYTEK
MT
GPT-4o-mini
TTS
iOS Voice
Latency
~1.5 s
VPN
Required
Cost
~$0.02 / min · ~$1.20 / h
Quality
★★★★☆

Required API keys

⚠ Not for: situations where GPT-4o quality is essential (use Pro Live) or where no VPN is available (use China-Only).

02Pro Live

Best available translation quality. GPT-4o flagship for translation, MiniMax for more natural voice. For high-stakes negotiations, contract discussions, or ambiguous technical content where quality matters more than cost.

ASR
iFLYTEK
MT
GPT-4o
TTS
MiniMax
Latency
~2.5 s
VPN
Required
Cost
~$0.10 / min · ~$6.00 / h
Quality
★★★★★

Required API keys

  • Required iFLYTEK — App ID + API Key + API Secret
  • Required OpenAI — API key (GPT-4o access)
  • Required MiniMax — Group ID + API key
⚠ Not for: cost-sensitive everyday use (use Standard) or situations where no VPN is available.

03Realtime

OpenAI's Realtime API handles speech recognition, translation and audio in a single connection — no separate iFLYTEK credentials needed. Lowest latency but highest cost. For situations where every half-second counts.

ASR + MT + TTS
OpenAI Realtime API
Latency
< 1 s
VPN
Required
Cost
~$0.15 / min · ~$9.00 / h
Quality
★★★★☆

Required API keys

  • Required OpenAI — API key with Realtime API access
⚠ Not for: longer meetings where costs accumulate quickly; China users without VPN.

04China-Only

All three components run on servers within China — no outbound traffic to OpenAI or other international providers. No VPN required. This makes it the only pipeline that works reliably from any location in mainland China. Also the cheapest option.

ASR
iFLYTEK
MT
DeepSeek
TTS
MiniMax
Latency
~2 s
VPN
Not needed
Cost
~$0.01 / min · ~$0.60 / h
Quality
★★★☆☆

Required API keys

  • Required iFLYTEK — App ID + API Key + API Secret
  • Required DeepSeek (platform.deepseek.com) — API key
  • Required MiniMax — Group ID + API key
⚠ Not for: use cases outside China where GPT-4o quality matters; users outside China with a VPN who want the best quality.

05Watch-Modus

Same stack as Standard, but the Apple Watch is the primary display. Translation bubbles appear on the Watch. TTS is off by default — battery and meeting-friendly. The iPhone stays active and shows everything; Watch is an additional surface, not a replacement.

ASR
iFLYTEK
MT
GPT-4o-mini
Display
Watch + iPhone
Latency
~1.5 s
VPN
Required
Cost
~$0.02 / min · ~$1.20 / h
Quality
★★★★☆

Required API keys

  • Required iFLYTEK — App ID + API Key + API Secret
  • Required OpenAI — API key (gpt-4o-mini)
⚠ Not for: users without an Apple Watch; users who need TTS output during the meeting (enable TTS manually in settings).

06Which pipeline for which situation?

  • In China, no VPN available
    China-Only
  • High-stakes negotiation, quality is critical
    Pro Live
  • Fast back-and-forth, every second counts
    Realtime
  • Full-day meeting, watching costs
    Standard
  • Discreet transcript on your wrist
    Watch-Modus
  • First time, unsure what to pick
    Standard
Set up your API keys →