SDK

2d-talk

The on-device AI communication runtime.Voice, video, and intelligence — entirely local. Entirely private.

iOS & Android|Offline-First|Sub-100ms Latency

AI that runs where your users are

2d-talk is a complete AI communication stack — language models, speech recognition, voice synthesis, and visual understanding — compiled to run natively on mobile devices. No cloud APIs. No network requirement. No compromises on privacy.

Every other solution sends your users' voices and faces to someone else's server. We built the technology to make that unnecessary. The result is faster, more private, and works in places where nothing else can.

Why 2d-talk

Built different. By design.

Zero Cloud Dependency

Every model runs on-device. No data leaves the phone. No round-trip latency. No server costs scaling with your user base.

Deterministic Reliability

Automatic resource management across memory, thermal, and battery thresholds. The SDK degrades gracefully — it never crashes, never stutters, never drops a conversation.

Ship in Days, Not Months

Single SDK integration. Configure once with a JSON model pack. No ML expertise required. Native iOS Swift Package and Android Kotlin support out of the box.

Frontier Model Intelligence

Adaptive model routing selects the optimal model size per-query based on device state. 1B for efficiency, 3B for complex reasoning — automatically, transparently.

Complete Privacy

Audio never leaves the device. Video frames are processed locally and discarded. Conversations are stored in an encrypted local database. Fully HIPAA and GDPR compatible by architecture.

Works Everywhere

Airplanes, rural areas, underground, privacy-critical environments. No network dependency means your AI features work 100% of the time, for 100% of your users.

Cloud AI is a compromise

Cloud-based AI APIs

User data transmitted to third-party servers
Latency dependent on network conditions
Fails completely without connectivity
Server costs scale linearly with users
Privacy compliance requires legal review
Vendor lock-in with proprietary APIs

2d-talk

All processing stays on the user's device
Sub-100ms latency — hardware speed, not network speed
Full functionality offline, anywhere, always
Fixed per-user cost regardless of usage volume
Private by architecture — HIPAA/GDPR compatible by default
Standard native SDK — iOS Swift Package, Android Kotlin
Pricing

Simple, per-user pricing

No API call counting. No token metering. One flat rate per active user.

Text + Image

Lite
$0.40per user / mo

On-device LLM inference and image understanding. No cloud. No latency.

1B parameter language model
Real-time image tagging & analysis
Conversation memory with auto-summarization
Offline-first — works without connectivity
Signed usage metering

Text + Image

Full
$0.60per user / mo

Upgraded models with adaptive routing. Peak intelligence on any device.

Up to 3B parameter model with burst routing
Advanced vision model (multi-frame analysis)
Intelligent model selection per-query
Extended context with KV cache management
Full metering & analytics dashboard

Voice

Lite
$0.80per user / mo

Natural voice conversations running entirely on the user's device.

Speech-to-text with voice activity detection
Real-time LLM response streaming
Natural text-to-speech synthesis
Echo cancellation & noise suppression
Sub-100ms interaction latency
Most Popular

Voice

Full
$1.00per user / mo

Production-grade voice AI with barge-in, persona control, and graceful degradation.

Everything in Voice Lite
Natural interruption handling (barge-in)
Customizable voice personas & DSP filters
Automatic thermal & memory management
Deterministic fallback — never crashes

Video

Lite
$1.50per user / mo

Voice AI plus real-time visual understanding from the device camera.

Everything in Voice Full
Real-time camera vision tagging
Visual context injection into conversation
Lightweight vision model routing
Battery-aware processing

Video

Full
$2.00per user / mo

The complete stack. Voice, vision, and intelligence — all on-device, all private.

Everything in Video Lite
Multi-model vision routing (adaptive quality)
Watchdog-gated inference under pressure
Full resource cascade management
Enterprise metering & compliance tooling

Volume discounts available for 10,000+ MAU. Contact us for enterprise pricing.

Ready to build with 2d-talk?

Native SDKs for iOS and Android. Single integration. Configure with a JSON model pack and ship AI-powered communication in days.