termux-tts 1.5.0 → 1.5.2

package/doc.config.yaml

@@ -1,595 +0,0 @@
-# ==============================================================================
-# AMEVA Documentation Configuration: Termux-TTS
-# Strict Compliance with AOSF-ENG-STD-2026-V1 & OpenSSF Standards
-# Single Source of Truth (SSOT)
-# ==============================================================================
-
-name: "termux-tts"
-display_name: "Termux-TTS"
-package_name_pypi: "termux-tts"
-package_name_npm: "termux-tts"
-version: "v1.5.0"
-release_name: "100% Native Vulkan Hardware GPU Pipeline & Zero-Silent-Fallback Architecture"
-license: "Apache-2.0"
-platform: "Android ARM64 / Qualcomm Adreno & ARM Mali Vulkan 1.3 / Linux"
-github_repo_url: "https://github.com/uno-km/termux-tts"
-
-custom_pages:
-  - slug: "vulkan-engineering-paper"
-    title_en: "Vulkan C++ Engineering Paper"
-    title_ko: "Vulkan C++ 네이티브 가속 논문"
-    file: "lib/tts/vulkan-engineering-paper.html"
-
-tagline_en: "Production-Grade 4-Tier On-Device Speech Synthesis Framework (Multilingual Neural Orchestrator, Zero-Dependency DSP Formant & C++ Acceleration)"
-tagline_ko: "모바일 및 엣지 환경을 위한 프로덕션급 4-Tier 온디바이스 음성 합성 프레임워크 (다국어 신경망 오케스트레이터, 무의존성 DSP 포먼트, C++ 가속 엔진)"
-
-why_challenge_en: "Constrained mobile edge environments frequently suffer from execution instability, excessive thermal throttling, and unpredictable runtime memory spikes when running conventional deep learning text-to-speech stacks. Heavyweight dependencies like full PyTorch or unoptimized runtime graph compilers exhaust mobile DRAM, while cross-language code-switching historically required multiple disconnected runtimes or heavy cloud APIs."
-why_challenge_ko: "제약된 모바일 엣지 환경에서는 과도한 메모리 점유, 발열 스로틀링, 드라이버 파편화로 인해 생성형 음성 합성 모델 구동이 불안정합니다. 기존 무거운 프레임워크는 수 기가바이트의 런타임 의존성으로 모바일 OOM을 유발하며, 다국어 교차 발화(한/영/일 등)는 단일 기기 로컬에서 매끄럽게 처리하기 어려웠습니다."
-
-description_en: "Termux-TTS delivers a resilient 4-Tier on-device text-to-speech architecture designed for deterministic latency, zero-config ergonomics, and multi-language acoustic fidelity. It bridges lightweight parametric DSP synthesis (<50ms compute, 0MB model download) with high-fidelity resident C-API multilingual neural models (9 languages including Korean, English, Japanese, Hindi, and Russian at 22.05kHz), alongside Android system native speech service routing. With automated 1-click provisioning, Unicode script classification, and resident in-memory model caching, Termux-TTS achieves high acoustic fidelity with sub-0.18x real-time factor."
-description_ko: "Termux-TTS는 결정론적 지연 시간, Zero-Config 사용성, 다국어 고품질 음향을 실현하는 프로덕션급 4-Tier 온디바이스 음성 합성 프레임워크입니다. 0MB 무의존성 파라메트릭 DSP 보코더부터 9개 공식 언어(한국어, 영어, 일본어, 힌디어, 러시아어 등)를 지원하는 레지던트 C-API 다국어 신경망 오케스트레이터, 그리고 안드로이드 네이티브 음성 브릿지까지 통합 제공합니다."
-
-quick_install_cmd: |
-  pip install termux-tts
-  # or: npm install termux-tts
-  # 1-Click Provision Default Models (Korean KSS + English Lessac):
-  termux-tts install
-  # On-demand provision specific languages (e.g. Hindi, Japanese, Russian, or all):
-  termux-tts install --models hi
-  # Instant speech synthesis:
-  termux-tts "Hello 방가방가 나는 parrot 이라고 해." -o out.wav --play
-
-features:
-  - title_en: "Multilingual Neural Orchestrator"
-    desc_en: "Dynamic cross-language code-switching and single-language synthesis supporting 9 official languages (Korean, English, Japanese, Chinese, Hindi, Russian, Spanish, French, German) with 50ms context-aware pause padding."
-    title_ko: "다국어 신경망 오케스트레이터"
-    desc_ko: "한국어, 영어, 일본어, 중국어, 힌디어, 러시아어, 스페인어, 프랑스어, 독일어 9대 언어의 다국어 교차 발화 및 단독 발화를 50ms 문맥 묵음 패딩과 함께 실시간 합성합니다."
-  - title_en: "Resident C-API In-Memory Acceleration"
-    desc_en: "Native C-API residency with SherpaResidentManager eliminates subprocess startup latency and achieves sub-0.18x real-time factor with ARM NEON SIMD acceleration."
-    title_ko: "레지던트 C-API 인메모리 가속"
-    desc_ko: "SherpaResidentManager를 통한 C-API 모델 인메모리 상주로 프로세스 기동 지연을 전면 배제하고 ARM NEON SIMD 가속으로 0.18x 이하의 RTF를 달성합니다."
-  - title_en: "Zero-Config CLI Ergonomics"
-    desc_en: "Top-level speech synthesis by default (termux-tts 'Hello' --play) without requiring subcommands or explicit engine flags. Explicit 'speak' command routes to Android native system voice."
-    title_ko: "Zero-Config CLI 사용성"
-    desc_ko: "서브 커맨드나 엔진 플래그 없이 문장을 직접 입력(termux-tts '문장' --play)하면 즉시 신경망 합성이 수행되며, 안드로이드 시스템 음성은 'speak'로 명시 분기합니다."
-  - title_en: "On-Demand Self-Healing Provisioner"
-    desc_en: "1-Click automated provisioning (termux-tts install --models default/hi/ja/ru/zh/all) with actionable English guidance and absolute paths for uninstalled models."
-    title_ko: "온디맨드 자가 치유 프로비저너"
-    desc_ko: "최초 설치 시 기본 모델(한/영)만 경량 배포하고, 미설치 언어 호출 시 절대 경로와 함께 즉시 복구 가능한 영문 안내문 및 온디맨드 설치 명령을 제공합니다."
-  - title_en: "4-Tier Resilient Architecture"
-    desc_en: "Tier 1: Zero-Dependency Parametric DSP Formant (0MB footprint). Tier 2: Android Native System Voice Bridge. Tier 3: Subprocess-Isolated Sherpa C++ CPU Engine. Tier 4: Pure Vulkan GPU Hardware Neural Acceleration."
-    title_ko: "4-Tier 복원형 합성 아키텍처"
-    desc_ko: "Tier 1 무의존성 DSP 포먼트, Tier 2 안드로이드 네이티브 음성 브릿지, Tier 3 격리형 Sherpa CPU 엔진, Tier 4 순수 Vulkan GPU 하드웨어 신경망 가속."
-  - title_en: "1-Click Automated Provisioning"
-    desc_en: "Command-line provisioning (termux-tts install --tier high/medium) automatically resolves precompiled ARM64 native binaries and HuggingFace weights with self-test verification."
-    title_ko: "1-클릭 자동 프로비저닝"
-    desc_ko: "명령어 1회로 ARM64 네이티브 바이너리와 허깅페이스 가중치를 자동 감지 및 설치하고 자체 진단 테스트를 완결합니다."
-  - title_en: "Studio-Grade FP16 Neural Fidelity"
-    desc_en: "Native support for high-resolution 22.05kHz 16-bit PCM studio models (vits-piper-en_US-lessac-high-fp16), delivering natural prosody on mobile hardware."
-    title_ko: "스튜디오급 FP16 고음질 신경망"
-    desc_ko: "22.05kHz 16비트 PCM 고해상도 레퍼런스 모델을 지원하여 극도로 자연스러운 억양과 음향 충실도를 모바일 단말기에서 제공합니다."
-  - title_en: "Pure Vulkan GPU Acceleration (Zero CPU Fallback)"
-    desc_en: "Direct GPU tensor compute via Vulkan 1.3 pipeline caching, fully resolving Qualcomm Adreno 830 and ARM Mali-G68 shader driver anomalies without silent CPU fallback."
-    title_ko: "순수 Vulkan GPU 가속 (Zero CPU 침묵 폴백)"
-    desc_ko: "침묵 폴백 없이 Qualcomm Adreno 830 및 ARM Mali-G68의 셰이더 드라이버 결함을 완전 해소하고 순수 GPU 텐서 연산을 수행합니다."
-  - title_en: "Subprocess IPC Memory Isolation"
-    desc_en: "Architectural separation of audio synthesis and OpenSL ES playback threads via subprocess IPC prevents Python GIL contention and memory corruption."
-    title_ko: "프로세스 격리형 메모리 보호"
-    desc_ko: "음성 합성 파이프라인과 오디오 재생 스레드를 서브프로세스 IPC로 분리하여 GIL 병목 및 메모리 오염을 원천 차단합니다."
-  - title_en: "Unified Cross-Language Toolchain"
-    desc_en: "Identical functional parity and strict typing across Global CLI, Python 3.10+ SDK, and Node.js / TypeScript runtime bindings."
-    title_ko: "통합 크로스 랭귀지 툴체인"
-    desc_ko: "Global CLI, Python 3.10+ SDK, Node.js / TypeScript 런타임 간 완전한 기능 동등성과 타입 안전성을 보장합니다."
-
-matrix_table:
-  - category: "Tier 1: DSP Formant"
-    operations: "Rosenberg Glottal Pulse, 5-Band Biquad Filter, Sino-Korean Numeral Normalizer"
-    status: "Production (<50ms)"
-  - category: "Tier 2: Native Bridge"
-    operations: "Termux-API / Android TTS Service IPC (Samsung/Google Engine)"
-    status: "Production (Immediate)"
-  - category: "Tier 3: CPU Neural"
-    operations: "Sherpa C++ Subprocess-Isolated VITS Engine (ARM64 NEON)"
-    status: "Production (RTF 0.4~1.2x)"
-  - category: "Tier 4: Vulkan GPU"
-    operations: "C++ Vulkan Pipeline Caching, Mali Quirk Handling, Adreno Subgroup 64"
-    status: "Production (RTF 0.26x)"
-
-code_example_py: |
-  import termux_tts as tts
-
-  # 1. Zero-Config Multilingual Neural Synthesis (Korean + English Code-Switching)
-  with tts.load() as engine:
-      result = engine.synthesize(
-          "Hello 방가방가 키키키키 나는 parrot 이라고 해. Natural cross-language neural voice.",
-          output="multilingual.wav"
-      )
-      print(f"Synthesized {result.duration_sec:.2f}s in {result.elapsed_ms:.1f}ms (RTF: {result.rtf:.4f}x)")
-
-  # 2. Pure Vulkan GPU Neural Synthesis (Studio Tier)
-  with tts.load(engine="vulkan", model_tier="high") as engine:
-      result = engine.synthesize(
-          "The neural speech synthesis engine is operating with pure Vulkan hardware acceleration.",
-          output="studio_vulkan.wav"
-      )
-      print(f"Synthesized {result.duration_sec:.2f}s in {result.elapsed_ms:.1f}ms (RTF: {result.rtf:.4f}x)")
-
-  # 3. Instant Zero-Dependency DSP Formant Synthesis
-  with tts.load(engine="dsp", preset="balanced") as engine:
-      result = engine.synthesize("Instant speech generation with zero external weights.", output="dsp.wav")
-      print(f"DSP Synthesis Latency: {result.elapsed_ms:.1f}ms")
-
-  # 4. Direct Android Hardware Speaker Playback
-  with tts.load(engine="native", language="en") as engine:
-      engine.speak("Direct hardware speaker output via Android native service.")
-
-code_example_js: |
-  const tts = require('termux-tts');
-
-  async function main() {
-    // 1. Initialize Vulkan GPU Neural Engine
-    const engine = tts.load({ engine: 'vulkan', tier: 'high' });
-    const res = await engine.synthesize(
-      "High-performance speech synthesis on mobile hardware.",
-      { output: "vulkan_output.wav" }
-    );
-    console.log(`Generated: ${res.durationSec}s in ${res.elapsedMs}ms (RTF: ${res.rtf}x)`);
-
-    // 2. Hardware Diagnostics
-    const diag = await tts.doctor();
-    console.log(`Vulkan GPU Device: ${diag.device_name} (API: ${diag.api_version})`);
-  }
-  main();
-
-benchmarks:
-  headers: ["Hardware Target", "SoC / Physical GPU Architecture", "Vulkan Driver ABI", "Audio Duration", "Synthesis Latency", "Real-Time Factor (RTF)", "Execution Status"]
-  rows:
-    - ["Galaxy S21", "Exynos 2100 / ARM Mali-G78 (0x9800000)", "Vulkan 1.1 (/system/lib64)", "4.25 s", "9.37 s", "2.2055x", "Validated (100% Native GPU)"]
-    - ["Galaxy S25", "Snapdragon 8 Elite / Adreno 830 (0x80320040)", "Vulkan 1.3 (/system/lib64)", "4.25 s", "16.21 s", "3.8145x", "Validated (100% Native GPU)"]
-    - ["Galaxy S22", "Snapdragon 8 Gen 1 / Adreno 730 (0x80267062)", "Vulkan 1.1 (/system/lib64)", "4.27 s", "38.09 s", "8.9159x", "Validated (100% Native GPU)"]
-    - ["Galaxy A35", "Exynos 1380 / ARM Mali-G68 MP5 (0x9801000)", "Vulkan 1.1 (/system/lib64)", "4.27 s", "51.91 s", "12.1505x", "Validated (100% Native GPU)"]
-    - ["Heterogeneous CPU", "Cortex-A78 / A55 CPU Core", "Sherpa NEON SIMD C++", "4.25 s", "1.85 s", "0.4350x", "Validated (CPU Reference)"]
-
-api_reference:
-  - symbol: "termux_tts.load(engine='auto'|'vulkan'|'sherpa'|'dsp'|'native', model_tier='high'|'medium', preset='balanced')"
-    description: "Factory initializer returning the configured synthesis engine context with resource finalization."
-  - symbol: "engine.synthesize(text: str, output: str = None, speed: float = 1.0, preset: str = None) -> SynthesisResult"
-    description: "Executes phonetic tokenization, neural/DSP acoustic modeling, and writes 16-bit PCM WAV audio."
-  - symbol: "engine.speak(text: str, stream: str = 'MUSIC') -> NativeResult"
-    description: "Routes speech synthesis directly to the physical device speakers via Android system native services."
-  - symbol: "termux_tts.run_installation(tier='high'|'medium', force=False) -> bool"
-    description: "Automated provisioning toolchain downloading native C++ binaries from GitHub Releases and HuggingFace weights."
-  - symbol: "termux_tts.doctor() -> DiagnosticReport"
-    description: "Runs rigorous 12-stage hardware diagnostics verifying Vulkan instance, GPU compute queues, and memory buffers."
-
-installation_body: |
-  <h3>Standard Package Installation</h3>
-  <p>Install the core Python or Node.js package directly from standard package repositories:</p>
-  <pre><code># Python (PyPI)
-  pip install termux-tts
-
-  # Node.js / TypeScript (NPM)
-  npm install termux-tts</code></pre>
-
-  <h3>Prerequisites on Android Termux</h3>
-  <p>Ensure that the required system packages are installed in your Termux environment:</p>
-  <pre><code>pkg update && pkg install -y termux-api pulseaudio sox clang</code></pre>
-
-  <h3>1-Click Automated Engine Provisioning</h3>
-  <p>To enable Tier 4 Vulkan GPU neural speech synthesis without manual compilation or model downloading, run the built-in installer:</p>
-  <pre><code># Install High-Resolution Studio Tier (vits-piper-en_US-lessac-high-fp16, 22.05kHz)
-  termux-tts install --tier high
-
-  # Or install Medium Compact Tier (vits-piper-en_US-lessac-medium)
-  termux-tts install --tier medium</code></pre>
-  <p>The provisioning script downloads the precompiled native ARM64 C++ binary (<code>sherpa-ncnn-offline-tts-vulkan</code>) from the official GitHub Release and verifies execution via an immediate self-test.</p>
-
-quickstart_body: |
-  <h3>Command-Line Interface (CLI) Recipes</h3>
-  <pre><code># 1. High-Resolution Studio Neural Synthesis with Speaker Playback
-  termux-tts synth -e vulkan --tier high -t "Operating at full hardware capacity." -o speech.wav --play
-
-  # 2. Instant Zero-Dependency DSP Formant Synthesis
-  termux-tts synth -e dsp -p ultra -t "Zero dependency parametric speech synthesis." -o dsp.wav
-
-  # 3. Direct Android Hardware Speaker Broadcast
-  termux-tts speak -t "System notification broadcast." -l en
-
-  # 4. Hardware and Environment Diagnostics
-  termux-tts doctor</code></pre>
-
-  <h3>Python SDK Canon</h3>
-  <pre><code>import termux_tts as tts
-
-  # Initialize pure Vulkan GPU neural synthesis engine
-  with tts.load(engine="vulkan", model_tier="high") as engine:
-      res = engine.synthesize("Validating deterministic tensor execution.", output="out.wav")
-      print(f"Elapsed: {res.elapsed_ms:.1f}ms | RTF: {res.rtf:.4f}x")</code></pre>
-
-api_body: |
-  <h3>Factory Gateway</h3>
-  <table class="data-table">
-    <thead>
-      <tr><th>Symbol</th><th>Signature</th><th>Description</th></tr>
-    </thead>
-    <tbody>
-      <tr>
-        <td><code>termux_tts.load</code></td>
-        <td><code>(engine: str = 'auto', model_tier: str = 'high', preset: str = 'balanced', language: str = 'en')</code></td>
-        <td>Initializes and returns the unified TTS engine instance. Supports context manager pattern.</td>
-      </tr>
-      <tr>
-        <td><code>termux_tts.run_installation</code></td>
-        <td><code>(tier: str = 'high', force: bool = False) -> bool</code></td>
-        <td>1-click provisioning tool downloading native ARM64 C++ binaries and model weights.</td>
-      </tr>
-      <tr>
-        <td><code>termux_tts.doctor</code></td>
-        <td><code>() -> DiagnosticReport</code></td>
-        <td>Executes comprehensive 12-stage Vulkan and system diagnostics.</td>
-      </tr>
-    </tbody>
-  </table>
-
-  <h3>Engine Methods</h3>
-  <table class="data-table">
-    <thead>
-      <tr><th>Method</th><th>Parameters</th><th>Return Type</th><th>Description</th></tr>
-    </thead>
-    <tbody>
-      <tr>
-        <td><code>synthesize</code></td>
-        <td><code>(text: str, output: str = None, speed: float = 1.0, preset: str = None)</code></td>
-        <td><code>SynthesisResult</code></td>
-        <td>Synthesizes text into 16-bit linear PCM WAV audio.</td>
-      </tr>
-      <tr>
-        <td><code>speak</code></td>
-        <td><code>(text: str, stream: str = 'MUSIC')</code></td>
-        <td><code>NativeResult</code></td>
-        <td>Transmits speech to the device speaker via Android native TTS bridge.</td>
-      </tr>
-    </tbody>
-  </table>
-
-benchmarks_body: |
-  <h3>Empirical Dual-Device Hardware Benchmark Suite</h3>
-  <p>The following performance benchmarks were gathered directly on physical test devices running Android 16 under Termux ARM64 environment. Every test was conducted over repeated synthesis iterations with zero silent fallback.</p>
-
-  <table class="data-table">
-    <thead>
-      <tr>
-        <th>Device &amp; Hardware Profile</th>
-        <th>Synthesis Engine</th>
-        <th>Model Architecture</th>
-        <th>Audio Duration</th>
-        <th>Compute Latency</th>
-        <th>Real-Time Factor</th>
-        <th>Hardware Utilization</th>
-      </tr>
-    </thead>
-    <tbody>
-      <tr>
-        <td><strong>Samsung Galaxy S25</strong><br>Snapdragon 8 Elite / Adreno 830</td>
-        <td>Vulkan GPU Neural</td>
-        <td>lessac-high-fp16 (22.05kHz)</td>
-        <td>6.70 s</td>
-        <td><strong>6.65 s</strong></td>
-        <td><strong>0.993x</strong></td>
-        <td>GPU Subgroup 64 / Pure Tensor</td>
-      </tr>
-      <tr>
-        <td><strong>Samsung Galaxy S25</strong><br>Snapdragon 8 Elite / Adreno 830</td>
-        <td>Vulkan GPU Neural</td>
-        <td>lessac-medium (22.05kHz)</td>
-        <td>4.59 s</td>
-        <td><strong>1.21 s</strong></td>
-        <td><strong>0.264x</strong></td>
-        <td>GPU Compute / 3.79x Faster Than RT</td>
-      </tr>
-      <tr>
-        <td><strong>Samsung Galaxy A35</strong><br>Exynos 1380 / ARM Mali-G68 MP5</td>
-        <td>Vulkan GPU Neural</td>
-        <td>lessac-medium (22.05kHz)</td>
-        <td>4.52 s</td>
-        <td><strong>5.18 s</strong></td>
-        <td><strong>1.146x</strong></td>
-        <td>Mali Subgroup 16 / Medium Kernel</td>
-      </tr>
-      <tr>
-        <td><strong>Samsung Galaxy A35</strong><br>Exynos 1380 / ARM Mali-G68 MP5</td>
-        <td>Vulkan GPU Neural</td>
-        <td>lessac-high-fp16 (22.05kHz)</td>
-        <td>6.73 s</td>
-        <td><strong>34.33 s</strong></td>
-        <td><strong>5.098x</strong></td>
-        <td>Full VRAM Resident / High Fidelity</td>
-      </tr>
-      <tr>
-        <td><strong>Heterogeneous ARM64</strong><br>All Core Profiles</td>
-        <td>Parametric DSP Formant</td>
-        <td>5-Band Biquad Formant</td>
-        <td>4.15 s</td>
-        <td><strong>0.054 s</strong></td>
-        <td><strong>0.0130x</strong></td>
-        <td>0MB External RAM / Instant CPU</td>
-      </tr>
-    </tbody>
-  </table>
-
-advanced_parameters_body: |
-  <h3>Architecture &amp; Hardware Driver Postmortem</h3>
-  <p>Operating deep learning inference models on mobile Vulkan runtimes involves navigating severe GPU driver irregularities. Below is the technical breakdown of bugs identified and permanently resolved in Termux-TTS v1.3.0:</p>
-
-  <h4>1. ARM Mali-G68 Valhall Subgroup Truncation Resolution</h4>
-  <p>On ARM Mali-G68 GPUs (subgroup size 16), unaligned matrix multiplication workgroups caused an integer division truncation: <code>loadstride_b = gl_WorkGroupSize.x * LOAD_VEC_B / BK = 16 * 1 / 32 = 0</code>. This resulted in an infinite loop (<code>for (uint l = 0; l &lt; BN; l += 0)</code>) leading to GPU command timeout and device loss (<code>VK_ERROR_DEVICE_LOST</code>). Termux-TTS enforces medium tile alignment and subgroup-aware shader dispatch, eliminating the loop and ensuring full 100% layer execution.</p>
-
-  <h4>2. Qualcomm Adreno 830 SPIR-V Pipeline Caching</h4>
-  <p>On Snapdragon 8 Elite Adreno 830 hardware, repeated runtime pipeline generation triggered JIT compiler crashes (<code>VK_ERROR_UNKNOWN -13</code>). Termux-TTS isolates model pipelines with immutable descriptor pooling and shader pipeline caching, stabilizing execution across long synthesis streams.</p>
-
-  <h4>3. Subprocess IPC Architecture for Audio Playback</h4>
-  <p>In-process audio playback via native libraries frequently triggered memory corruption and GIL deadlocks on Android Bionic. Termux-TTS orchestrates playback using isolated subprocess worker pools, safeguarding the primary application runtime.</p>
-
-changelog:
-  - version: "v1.4.4"
-    date: "2026-09-14"
-    title: "Multilingual Neural Orchestrator & Zero-Config Ergonomics"
-    type: "Production Release (Latest)"
-    changes:
-      - "Implemented MultilingualNeuralEngine supporting dynamic cross-language code-switching across 9 languages (ko, en, ja, zh, hi, ru, es, fr, de)."
-      - "Built SherpaResidentManager for native in-memory C-API residency with sub-0.18x RTF and zero subprocess startup lag."
-      - "Promoted speech synthesis to default top-level subcommand, accepting positional text without '-t' or '-e multilingual'."
-      - "Integrated universal Unicode script classifier covering Hangul, Latin, Devanagari, Cyrillic, CJK, and Arabic scripts."
-      - "Enhanced 1-click installer with lightweight default models (ko+en) and on-demand provisioning (--models <lang> / all)."
-      - "Hardened defensive path handling with Path.expanduser() across audio exports and CLI arguments."
-
-  - version: "v1.3.0"
-    date: "2026-09-05"
-    title: "Production 4-Tier Synthesizer & Studio Vulkan GPU Acceleration"
-    type: "Stable Release"
-    changes:
-      - "Implemented Tier 4 Pure Vulkan GPU Neural Engine (sherpa-ncnn-offline-tts-vulkan) with zero CPU fallback."
-      - "Integrated studio-grade reference model (vits-piper-en_US-lessac-high-fp16, 22.05kHz 16-bit PCM)."
-      - "Implemented 1-click automated provisioning toolchain (termux-tts install --tier high/medium) downloading ARM64 native binaries from GitHub Release."
-      - "Resolved ARM Mali-G68 subgroup size 16 shader integer truncation and Adreno 830 pipeline cache crashes."
-      - "Achieved empirical RTF 0.264x on Galaxy S25 Adreno 830 and RTF 1.146x on Galaxy A35 Mali-G68."
-      - "Unified cross-platform CLI, Python SDK, and Node.js / TypeScript runtime bindings."
-
-  - version: "v1.1.5"
-    date: "2026-09-04"
-    title: "Subprocess Isolation & Conversational Tag Expansion"
-    type: "Archive Release"
-    changes:
-      - "Implemented subprocess IPC audio playback isolation."
-      - "Added expressive conversational token handling."
-      - "Stabilized Termux-API native bridge integration."
-
-readme_content: |
-  # Termux-TTS
-
-  [![PyPI](https://img.shields.io/pypi/v/termux-tts.svg?style=flat-square&color=0369a1)](https://pypi.org/project/termux-tts/)
-  [![Python](https://img.shields.io/pypi/pyversions/termux-tts.svg?style=flat-square)](https://pypi.org/project/termux-tts/)
-  [![npm](https://img.shields.io/npm/v/termux-tts.svg?style=flat-square&color=b91c1c)](https://www.npmjs.com/package/termux-tts)
-  [![License](https://img.shields.io/badge/License-Apache_2.0-004499.svg?style=flat-square)](https://github.com/uno-km/termux-tts)
-
-  > Production-Grade 4-Tier On-Device Speech Synthesis Framework (Multilingual Neural Orchestrator, Zero-Dependency DSP Formant & C++ Acceleration)
-
-  ---
-
-  ## Architecture & Overview
-
-  Termux-TTS is an enterprise-grade, on-device text-to-speech framework optimized for mobile edge hardware and Android Termux environments. Built to eliminate heavy dependency stacks and fragile driver behaviors, it features a resilient 4-Tier architecture:
-
-  - **Multilingual Neural Orchestrator**: Dynamic cross-language code-switching mesh covering 9 official languages (Korean, English, Japanese, Chinese, Hindi, Russian, Spanish, French, German) with Unicode script tokenization and 50ms context-aware silence padding.
-  - **Resident C-API In-Memory Engine**: Direct C-API memory residency with `SherpaResidentManager` achieving sub-0.18x real-time factor with ARM NEON SIMD acceleration and zero subprocess lag.
-  - **Tier 1: Zero-Dependency Parametric DSP Formant Vocoder**: 0MB disk footprint, Rosenberg glottal pulse formulation, and 5-band biquad formant filters providing deterministic speech synthesis in under 50 milliseconds (RTF 0.013x).
-  - **Tier 2: Android System Native Voice Bridge**: Direct IPC integration to physical Samsung and Google speech engines via the Termux-API service layer.
-  - **Tier 3: Subprocess-Isolated Sherpa C++ CPU Engine**: Subprocess-isolated VITS acoustic modeling on ARM64 NEON with memory leak protection.
-  - **Tier 4: Pure Vulkan GPU Hardware Neural Engine**: High-performance GPU tensor synthesis via precompiled native C++ binaries (`sherpa-ncnn-offline-tts-vulkan`) running high-resolution studio models (`vits-piper-en_US-lessac-high-fp16`) with zero silent CPU fallback.
-
-  ---
-
-  ## Empirical Hardware Benchmarks (Physical Devices)
-
-  Measurements gathered on physical Android 16 hardware running Termux ARM64:
-
-  | Target Device | Hardware Architecture | Synthesis Engine | Model Profile | Audio Length | Synthesis Time | Real-Time Factor (RTF) | Status |
-  | :--- | :--- | :--- | :--- | :--- | :---: | :---: | :---: |
-  | **Galaxy A53** | Exynos 1280 / ARM64 NEON | Multilingual Neural C-API | `KSS + Lessac` (Bilingual) | 10.50 s | **1.82 s** | **0.173x** | Validated (5.78x RT) |
-  | **Galaxy S25** | Snapdragon 8 Elite / Adreno 830 | Vulkan GPU Neural | `lessac-high-fp16` | 6.70 s | **6.65 s** | **0.993x** | Validated |
-  | **Galaxy S25** | Snapdragon 8 Elite / Adreno 830 | Vulkan GPU Neural | `lessac-medium` | 4.59 s | **1.21 s** | **0.264x** | Validated (3.79x RT) |
-  | **Galaxy A35** | Exynos 1380 / Mali-G68 MP5 | Vulkan GPU Neural | `lessac-medium` | 4.52 s | **5.18 s** | **1.146x** | Validated |
-  | **ARM64 CPU** | Cortex-A78 / A55 | Parametric DSP | 5-Band Biquad | 4.15 s | **0.054 s** | **0.0130x** | Validated (76x RT) |
-
-  ---
-
-  ## Installation & Automated Provisioning
-
-  ### 1. Package Installation
-  ```bash
-  # Python SDK & CLI
-  pip install termux-tts
-
-  # Node.js / TypeScript
-  npm install termux-tts
-  ```
-
-  ### 2. Automated Model Provisioning
-  Automate downloading and linking precompiled models to the official immutable path (`/data/data/com.termux/files/home/models/tts/`):
-  ```bash
-  # Default lightweight installation (Korean KSS + English Lessac)
-  termux-tts install
-
-  # On-demand provision specific language models:
-  termux-tts install --models hi    # Hindi (Piper Swara)
-  termux-tts install --models ja    # Japanese (Piper Hina)
-  termux-tts install --models ru    # Russian (Piper Dmitri)
-  termux-tts install --models zh    # Chinese (AISHELL3)
-  termux-tts install --models all   # All 9 official languages
-
-  # Provision Studio Vulkan GPU engine:
-  termux-tts install --tier high
-  ```
-
-  ---
-
-  ## Quickstart
-
-  ### Global Command-Line Interface (Zero-Config)
-  ```bash
-  # 1. Zero-config synthesis with physical speaker playback
-  termux-tts "Hello 방가방가 나는 parrot 이라고 해." -o ~/out.wav --play
-
-  # 2. Flag-based syntax
-  termux-tts -t "Multilingual neural speech synthesis on device." --play
-
-  # 3. Force language pinning
-  termux-tts -l en -t "Pure English text output." --play
-  termux-tts -l ko -t "한국어 단독 신경망 음성 합성." --play
-
-  # 4. Instant DSP Formant synthesis (0MB footprint)
-  termux-tts synth -e dsp -t "Zero dependency DSP synthesis." -o dsp.wav
-
-  # 5. Direct Android system native speaker broadcast
-  termux-tts speak -t "Hardware speaker broadcast via Android service."
-
-  # 6. Hardware diagnostics
-  termux-tts doctor
-  ```
-
-  ### Python SDK
-  ```python
-  import termux_tts as tts
-
-  # 1. Zero-Config Multilingual Neural Synthesis (Korean + English Code-Switching)
-  with tts.load() as engine:
-      result = engine.synthesize(
-          "Hello 방가방가 키키키키 나는 parrot 이라고 해.",
-          output="multilingual.wav"
-      )
-      print(f"Generated {result.duration_sec:.2f}s in {result.elapsed_ms:.1f}ms (RTF: {result.rtf:.4f}x)")
-
-  # 2. Pure Vulkan GPU Neural Synthesis (Studio Tier)
-  with tts.load(engine="vulkan", model_tier="high") as engine:
-      result = engine.synthesize("Pure Vulkan neural execution on mobile.", output="speech.wav")
-      print(f"Synthesized in {result.elapsed_ms:.1f}ms (RTF: {result.rtf:.4f}x)")
-
-  # 3. Zero-Dependency DSP Formant Synthesis
-  with tts.load(engine="dsp", preset="balanced") as engine:
-      result = engine.synthesize("Instant speech without model downloads.", output="dsp.wav")
-      print(f"DSP Latency: {result.elapsed_ms:.1f}ms")
-  ```
-
-  ### Node.js / TypeScript
-  ```typescript
-  import * as tts from 'termux-tts';
-
-  async function main() {
-    const engine = tts.load({ engine: 'vulkan', tier: 'high' });
-    const res = await engine.synthesize("High performance speech synthesis.", { output: "speech.wav" });
-    console.log(`Synthesized in ${res.elapsedMs}ms`);
-  }
-  main();
-  ```
-
-  ---
-
-  ## Official Documentation & Benchmarks
-  - [Official Architecture & API Reference](https://uno-km.vercel.app/lib/tts/)
-  - [Ecosystem Metrics & Registry Stats](https://uno-km.vercel.app/foundation/metrics)
-  - [AMEVA Open-Source Foundation Portal](https://uno-km.vercel.app/foundation/index.html)
-
-  ---
-
-  ## License
-  Licensed under the Apache-2.0 License. Copyright (c) 2026 Eunho Kim ([@uno-km](https://github.com/uno-km)).
-
-readme_pypi_content: |
-  # Termux-TTS (Python)
-
-  [![PyPI](https://img.shields.io/pypi/v/termux-tts.svg?style=flat-square&color=0369a1)](https://pypi.org/project/termux-tts/)
-  [![Python](https://img.shields.io/pypi/pyversions/termux-tts.svg?style=flat-square)](https://pypi.org/project/termux-tts/)
-  [![License](https://img.shields.io/badge/License-Apache_2.0-004499.svg?style=flat-square)](https://github.com/uno-km/termux-tts)
-
-  > Production-Grade 4-Tier On-Device Speech Synthesis Framework for Mobile & Edge (Zero-Dependency DSP Formant, C++ Vulkan GPU Neural Engine & Android Native Voice Bridge)
-
-  ## Installation
-
-  ```bash
-  pip install termux-tts
-  ```
-
-  ### 1-Click Automated Engine Provisioning
-  ```bash
-  termux-tts install --tier high
-  ```
-
-  ## Quickstart
-
-  ```python
-  import termux_tts as tts
-
-  # 1. Studio Vulkan GPU Neural Engine
-  with tts.load(engine="vulkan", model_tier="high") as engine:
-      result = engine.synthesize("Neural speech synthesis on mobile GPU.", output="speech.wav")
-      print(f"Elapsed: {result.elapsed_ms:.1f}ms (RTF: {result.rtf:.4f}x)")
-
-  # 2. Zero-Dependency DSP Formant Mode
-  with tts.load(engine="dsp", preset="balanced") as engine:
-      result = engine.synthesize("Instant speech generation.", output="dsp.wav")
-      print(f"DSP Latency: {result.elapsed_ms:.1f}ms")
-
-  # 3. Direct Android Native Speaker Output
-  with tts.load(engine="native", language="en") as engine:
-      engine.speak("Direct hardware speaker output.")
-  ```
-
-  ## Benchmarks (Physical Devices)
-
-  | Target Device | Hardware Architecture | Synthesis Engine | Real-Time Factor (RTF) | Status |
-  | :--- | :--- | :--- | :---: | :---: |
-  | **Galaxy S25** | Snapdragon 8 Elite / Adreno 830 | Vulkan GPU (`lessac-high-fp16`) | **0.993x** | Validated |
-  | **Galaxy S25** | Snapdragon 8 Elite / Adreno 830 | Vulkan GPU (`lessac-medium`) | **0.264x** | Validated |
-  | **Galaxy A35** | Exynos 1380 / Mali-G68 MP5 | Vulkan GPU (`lessac-medium`) | **1.146x** | Validated |
-  | **ARM64 CPU** | All Core Profiles | Parametric DSP Formant | **0.0130x** | Validated |
-
-  ## Documentation
-  - [Official Documentation & API Reference](https://uno-km.vercel.app/lib/tts/)
-  - [GitHub Repository](https://github.com/uno-km/termux-tts)
-
-  ## License
-  Apache-2.0 License. Copyright (c) 2026 Eunho Kim (@uno-km).
-
-readme_npm_content: |
-  # Termux-TTS (Node.js & TypeScript)
-
-  [![npm](https://img.shields.io/npm/v/termux-tts.svg?style=flat-square&color=b91c1c)](https://www.npmjs.com/package/termux-tts)
-  [![License](https://img.shields.io/badge/License-Apache_2.0-004499.svg?style=flat-square)](https://github.com/uno-km/termux-tts)
-
-  > Production-Grade 4-Tier On-Device Speech Synthesis Framework for Mobile & Edge (Zero-Dependency DSP Formant, C++ Vulkan GPU Neural Engine & Android Native Voice Bridge)
-
-  ## Installation
-
-  ```bash
-  npm install termux-tts
-  ```
-
-  ## Quickstart
-
-  ```typescript
-  import * as tts from 'termux-tts';
-
-  async function main() {
-    const engine = tts.load({ engine: 'vulkan', tier: 'high' });
-    const res = await engine.synthesize("High performance speech synthesis on mobile.", { output: "output.wav" });
-    console.log(`Synthesized: ${res.durationSec}s in ${res.elapsedMs}ms (RTF: ${res.rtf}x)`);
-  }
-  main();
-  ```
-
-  ## Documentation
-  - [Official Documentation & API Reference](https://uno-km.vercel.app/lib/tts/)
-  - [GitHub Repository](https://github.com/uno-km/termux-tts)
-
-  ## License
-  Apache-2.0 License. Copyright (c) 2026 Eunho Kim (@uno-km).

package/docs/benchmarks.md

@@ -1,12 +0,0 @@
-# 모바일 지연시간 벤치마크 (Mobile Benchmarks)
-
-## 1. 실기기 벤치마크 결과
-
-| 디바이스 | AP 프로세서 | 모델 | RTF (Real-Time Factor) | 5초 문장 합성 소요시간 | 메모리 점유율 |
-| :--- | :--- | :--- | :--- | :--- | :--- |
-| **Galaxy S25 (SM-S931N)** | Snapdragon 8 Elite | VITS ONNX | **0.035x** | **0.18초** | ~68MB |
-| **Galaxy A35 (SM-A356N)** | Exynos 1380 | VITS ONNX | **0.118x** | **0.59초** | ~72MB |
-| **Galaxy S20 (SM-G981N)** | Snapdragon 865 | VITS ONNX | **0.142x** | **0.71초** | ~75MB |
-
-> **RTF 공식**: $\text{RTF} = \frac{\text{합성에 소요된 연산 시간 (초)}}{\text{생성된 오디오 길이 (초)}}$  
-> RTF가 1.0 미만이면 실시간 발화 속도보다 빠르게 합성됨을 의미합니다.