Skip to content

Services & Model Ecosystem Overview (BYOK Architecture) ​

DuRT adheres to a philosophy of pragmatism, absolute privacy, and transparent pricing. DuRT operates on a pure client-side BYOK (Bring Your Own Key) architecture: all cloud API requests originate directly from your Mac to official providers. No intermediate developer proxies, zero markup, and zero data interception.

Furthermore, DuRT leverages Apple Silicon hardware to provide 100% on-device offline models (Apple Native Speech and Translation), enabling full production workflows even without an Internet connection.


4 Core Capability Ecosystems ​

mermaid
flowchart TD
    subgraph DurtSettings["DuRT Settings Architecture"]
        direction TB
        C1["🎙️ Speech Recognition (ASR)<br>【Realtime Streaming vs File Batch】"]
        C2["🧠 Language Models & Translation (LLM)<br>【macOS 15 Offline Translation + 15+ LLMs】"]
        C3["🗣️ Text-to-Speech (TTS)<br>【Voice Design, Voice Cloning & Aggregators】"]
        C4["📦 Knowledge & Delivery Destinations<br>【Prompts, Glossaries & 6 Cloud Channels】"]
    end

    C1 --> M1["[View ASR Matrix](/services/asr-overview)"]
    C2 --> M2["[View LLM Matrix](/services/llm-overview)"]
    C3 --> M3["[View TTS Matrix](/services/tts-overview)"]
    C4 --> M4["[View Integrations Overview](/integrations/)"]

1. Speech Recognition (ASR) ​

  • Realtime Speech: Live desktop floating captions and dictation powered by Apple Native Speech (0 latency / 0 data), OneASR Streaming, and OpenAI Live;
  • File Transcription: Media file chunking and online URL subtitle extraction powered by OneASR (YouTube/Bilibili streaming URL ingest, millisecond timestamps, speaker diarization) and OpenAI Whisper-1;
  • Note: Apple Native Speech is strictly limited to streaming audio and does not support file transcription.
  • 🔗 Explore: Speech Recognition Overview & Matrix

2. Large Language Models & Translation (LLM) ​

  • On-Device Translation: macOS 15+ Apple Translation framework for 100% offline real-time bilingual subtitles;
  • 15+ Direct Commercial & Open-Source Models:
    • Global Flagships: OpenAI (GPT-4o), DeepSeek (V3/R1), Claude 3.7, Gemini 2.5, Grok 3;
    • Domestic Compliant: Alibaba Qwen, Zhipu GLM, Moonshot Kimi, Tencent Hunyuan, Baidu ERNIE;
    • Local Private: Ollama (100% offline local inference);
    • Fast Inference: SiliconFlow, Groq, OpenRouter.
  • 🔗 Explore: LLM & Translation Overview & Matrix

3. Text-to-Speech (TTS) ​

  • Xiaomi MiMo: Lifelike neural voices with Voice Design and Voice Cloning;
  • OpenAI TTS: 6 standard high-definition studio voices;
  • OneTTS Gateway: Open-source aggregator for EdgeTTS and F5-TTS cloning.
  • 🔗 Explore: TTS Overview & Matrix

4. Knowledge Assets & 6 Delivery Destinations ​

Crafted with precision for macOS.