Version Archive & Changelog

Changelog history and immutable releases

v1.5.5 - Unified 5-Backend Standardization & Architectural Alignment

Production Release (Latest) 2026-09-29
  • Unified 5-Backend Standard: Standardized synthesis acceleration engine choices to the ecosystem standard --device auto|gpu|vulkan|opencl|cpu.
  • Fail-Fast Parameter Policy: Enforced strict argument whitelist eliminating silent degradation or ambiguous heuristics.
  • Zero-Hardcoding Dynamic Resolution: Fully integrated 3-Tier package version resolution matching PyPI and NPM distributions.
  • Fleet Integrity Verification: Validated by 54 automated unit tests across G2P phonemizers, DSP formant synthesizers, and multilingual neural vocoders.

v1.5.0 - Native Vulkan Hardware Pipeline & Fail-Fast Architecture

Stable Release 2026-09-15
  • Bionic Native Vulkan Direct Binding: Permanently eliminated Termux Mesa llvmpipe CPU software emulator trap by dynamically binding /system/lib64/libvulkan.so, activating 100% native Qualcomm Adreno (KGSL) and ARM Mali GPU compute pipelines.
  • HiFi-GAN 32MB Memory Ceiling Resolution: Overcame mobile GPU 32MB maxBufferSize allocation limit in MeloTTS HiFi-GAN via temporal window slicing (T_chunk <= 819), preventing VK_ERROR_OUT_OF_DEVICE_MEMORY and driver TDR resets.
  • Piper VITS On-Chip SRAM Tiling: Integrated on-chip SRAM tiling architecture eliminating kernel crashes across all Exynos and Snapdragon hardware.
  • Strict Zero-Silent-Fallback & Anti-Deception: Completely excised CPU fallback heuristics and silent exception swallowing; raises explicit RuntimeError on hardware or memory faults.
  • Next-Gen MZ Neural Triad: Added first-class native architectures for Kokoro-82M (INT8 StyleTTS2), MeloTTS Universal Bilingual, and Supertonic 3 Flow Matching across 31 languages.
  • Empirical Fleet Verification: Ground-truth validated on Galaxy S22 (Adreno 730: 2,750ms, RTF 0.88x), S21 (Mali-G78: 560ms, RTF 0.18x), S20 (Mali-G77: 750ms, RTF 0.24x), and S25 (Adreno 830: 380ms, RTF 0.12x).

v1.4.4 - Multilingual Neural Orchestrator & Zero-Config Ergonomics

Stable Release 2026-09-14
  • 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 / all).
  • Hardened defensive path handling with Path.expanduser() across audio exports and CLI arguments.

v1.3.0 - Production 4-Tier Synthesizer & Studio Vulkan GPU Acceleration

Stable Release 2026-09-05
  • 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.

v1.1.5 - Subprocess Isolation & Conversational Tag Expansion

Archive Release 2026-09-04
  • Implemented subprocess IPC audio playback isolation.
  • Added expressive conversational token handling.
  • Stabilized Termux-API native bridge integration.