About CrossMeet

Built for people who read 80%, miss the rest

A real engineering pain turned into a product that helps everyone.

Origin

"I can't really listen well, and my spoken English is rough. Sometimes I miss what the client says; sometimes I want to express something but freeze. But I can read most words just fine."

"If ASR shows me the source text, I instantly know what they said. The translation may lag a bit, but I already understand 80% from the English text. I just glance at the translation later to fill in the rest."

"Therefore, the #1 priority is putting ASR-recognized source text on screen as fast as possible."

— Todd, Founder

Timeline

  • 2025 · Fall — First overseas client meeting, missed key words, decided to write a prototype
  • 2025 · Winter — V1 working: local Whisper + ChatGPT translation, source text in 100ms
  • 2026 · Spring — Bidirectional simul-interpret + virtual mic + compose bar
  • 2026 · April — Internal dev: RAG knowledge base + license backend + glossary + quick phrases
  • 2026 · Mayv1.0.0 stable release (Windows)

Design principles

Iron Rule #1 — Source text always reaches the UI before translation.

The moment ASR has a token, the front-end shows it. The LLM translation, no matter how slow, never blocks this stream. This is the single biggest experiential difference between CrossMeet and other tools.

Iron Rule #2 — LLM never blocks ASR.

Multiple LLM calls run in parallel; results back-fill into existing source slots. Waiting for translation never becomes waiting for speech.

Every user-perceptible parameter is configurable.

VAD threshold, partial frequency, segmentation silence, dedup window — all exposed in Settings, because noisy meeting rooms and quiet offices need different optima.

Product philosophy

CrossMeet is "a tool for people who can't catch / produce English in real-time conversations" — not a generic translator. Under that focus, we don't:

  • Build flashy AI assistant chats
  • Build multi-party meeting summaries (Otter / Granola already do)
  • Build full simul-interpret (way beyond translation quality alone)

We do exactly three things:

  1. Listen — ASR source text in 100ms
  2. Speak — native-language input → TTS translation → virtual mic to the other side
  3. Simul-interpret — for business meetings, manually correct before sending. Zero translation accidents.

Corporate entities

CrossMeet is jointly operated by two legal entities. They are compliance-independent, invoice-independent, and book-independent — yet share a single product engineering team and a single codebase.

[CN_COMPANY_NAME] · China mainland

  • Unified Social Credit Code: [USCC]
  • Registered: [CN_CITY]
  • Scope: Software development, technical services, technical consulting (per business license)
  • Handles: China mainland sales, RMB settlement, VAT special/general invoices, PIPL compliance, ICP filing
  • Contact: sales-cn@crossmeet.com · billing-cn@crossmeet.com

[HK_COMPANY_NAME] · Hong Kong SAR

  • Business Registration Number: [BRN]
  • Certificate of Incorporation: [CR]
  • Registered Office: Hong Kong SAR
  • Handles: International sales, USD / HKD settlement (Stripe / PayPal / Wire Transfer), GDPR data processing, international invoicing
  • Contact: sales@crossmeet.com · dpo@crossmeet.com

Data processing & applicable law

  • Account and payment data for China mainland users are processed by [CN_COMPANY_NAME], stored within China mainland, and governed by the laws of the People's Republic of China.
  • Account and payment data for users outside China mainland are processed by [HK_COMPANY_NAME], stored in Hong Kong / overseas, and governed by the laws of the Hong Kong SAR.
  • For detailed data flow and your rights, see /privacy. For contracting entity and dispute resolution, see /terms.

Team

CrossMeet is built by a founder-led, focused product engineering team. No marketing department, no sales pressure, no investor roadmap to satisfy. Every line of code, every prompt, every support email is directly accountable to the product.

Contact

Get started

Ready to make every conversation feel native?

NO CREDIT CARD CANCEL ANYTIME LOCAL-FIRST
~100ms
End-to-end latency
30+
Languages
4
ASR engines
WIN 10/11
Native platform