Collaborative Sessions: Real-Time Translation Between Languages

Imagine sitting across from someone who speaks a different language. No interpreter, no phone switching, no copy-pasting. You both just talk. dijin transcribes and translates in real-time — each person reads in their own language.

🔑

dijin detects two different languages automatically and enables bidirectional translation — no settings to toggle, no buttons to press.

Two People, One Device, Two Languages

When two speakers use different languages on the same device, dijin activates cross-translation mode automatically. The system detects two distinct languages and enables bidirectional translation on its own.

How Cross-Translation Activates

First Speaker Talks (Japanese)

The speech engine transcribes, diarization assigns "speaker_0." System records: speaker_0 speaks Japanese. One language detected — cross-translation stays off.

Second Speaker Talks (Turkish)

The speech engine transcribes, diarization assigns "speaker_1." System records: speaker_1 speaks Turkish. Two languages detected — cross-translation activates automatically.

Bidirectional Translation Begins

Every segment from speaker_0 gets translated to Turkish, every segment from speaker_1 gets translated to Japanese. Automatic, continuous, real-time.

Two Translation Layers

dijin maintains two independent translation streams that never interfere with each other:

✗ Target Language Translation

✗Your chosen translation language

✗All non-target segments get translated

✗Works the same as single-speaker mode

✗Completely unchanged from standard mode

✓ Cross-Translation

✓Each speaker's text translated to the other's native language

✓Only appears when two different languages are detected

✓Automatic bidirectional

✓New layer added on top of existing

What You See on Screen

Cross-Translation Example

Speaker 1 (JA): こんにちは、会議について話しましょう

→ Target (TR): Toplantı hakkında konuşalım

⇄ Cross (TR): Toplantı hakkında konuşalım

Speaker 2 (TR): Evet, gündem maddelerini geçelim

→ Target: (same language — skipped)

⇄ Cross (JA): はい、議題を確認しましょう```

Speaker Memory Across Days

ℹ

dijin's diarization remembers speakers across sessions. Conversation with Yuki on Monday, meet again Thursday — dijin recognizes Yuki automatically. The speaker profile carries language preference, so cross-translation activates instantly without re-detection.

Works Offline

Apple Translation powers the on-device translation engine. No internet required for the core translation flow. Your audio and translations stay entirely on your device.

Online Multi-Participant Sessions

Beyond local two-person mode, dijin supports online collaborative sessions:

Feature	How It Works
Session Creation	Host creates with join code + PIN
Participant Joining	From their own devices
Real-time Transcript	Broadcasted via WebSocket
Language Preference	Each participant sees their language
Rate Limiting	200ms interval, 500 char cap

Local + Online Combined

The two-person local mode and online multi-participant mode work together. Two people sharing a device can host an online session where remote participants also join — everyone sees the transcript in their own language.

Use Cases

International meetings

Business

Patient-doctor communication

Medical

Local conversations

Travel

Multi-language classrooms

Education

The collaborative session infrastructure is live. The cross-translation local mode is in active development.