v0.0.10

SpeechCore XCFramework

Checksum: f1a4219300e9d26644fc8b2aea60e4fa21b346fca2ca76e202f6b2a4b8c4870f

Usage in Package.swift

.binaryTarget(
    name: "CSpeechCore",
    url: "https://github.com/soniqo/speech-core/releases/download/v0.0.10/SpeechCore.xcframework.zip",
    checksum: "f1a4219300e9d26644fc8b2aea60e4fa21b346fca2ca76e202f6b2a4b8c4870f"
)

v0.0.9

Highlights

Streaming speech recognition

• Added the low-memory Parakeet-EOU 120M ONNX runtime for streaming transcription with explicit end-of-utterance tokens.
• Expanded RNN-T/TDT decoding with opt-in beam search, contextual phrase biasing, an over-bias cap, and an optional emit reward. The default emit reward is zero, preserving previous decoding behavior.
• Added native Whisper ONNX support from small through large-v3/turbo, including automatic or fixed-language prompts, profiling, and CPU tuning controls.
• Parakeet now relies on model language autodetection; obsolete language-forcing behavior and documentation were removed.

Speech synthesis and voice cloning

• Added or expanded VoxCPM and VoxCPM2 ONNX runtimes, including split bundles, transcript-guided cloning, prompt/tokenizer fixes, and shared synthesis controls.
• Added CosyVoice3 ONNX integration with timing and decode controls.
• Added LiteRT runtimes for Supertonic-3, Chatterbox, and Indic-Mio.
• Added buffered TTS post-processing, spectral de-essing, and long-text chunking, with Kokoro improvements for short conversational turns.

Runtime, pipeline, and packaging

• Added FunctionGemma 270M through LiteRT-LM for local structured tool calls.
• Parallelized Hugging Face bundle downloads and kept the pre-speech ring active through the post-playback guard for faster turn-taking.
• Added tested Linux packages for amd64 and arm64, a documented speech command, clearer architecture-specific diagnostics, and a VoxCPM2 model downloader.
• Reworked the project README, added 13 translated mirrors, and linked Linux, Windows, Speech Core, and model guides on soniqo.audio.

Linux packages

Ubuntu 22.04+, Ubuntu 24.04+, and Debian 12+ are supported. The amd64 package includes ONNX and LiteRT CLI tools; arm64 is ONNX-only.

VERSION=0.0.9
ARCH=amd64 # or arm64
curl -fLO "https://github.com/soniqo/speech-core/releases/download/v${VERSION}/speech_${VERSION}_${ARCH}.deb"
sudo apt install "./speech_${VERSION}_${ARCH}.deb"
speech --help

See the Linux CLI reference and full comparison.

SpeechCore XCFramework

Checksum: 7bce60d25033b44288cf0e7b05eb79180f44c0b99db5654e5069f15893b71e42

Usage in Package.swift

.binaryTarget(
    name: "CSpeechCore",
    url: "https://github.com/soniqo/speech-core/releases/download/v0.0.9/SpeechCore.xcframework.zip",
    checksum: "7bce60d25033b44288cf0e7b05eb79180f44c0b99db5654e5069f15893b71e42"
)

v0.0.8

SpeechCore XCFramework

Checksum: d3fc8abc2d628307f78dd563211c37d7e87c7c90161dead6317d2367df4ba90f

Usage in Package.swift

.binaryTarget(
    name: "CSpeechCore",
    url: "https://github.com/soniqo/speech-core/releases/download/v0.0.8/SpeechCore.xcframework.zip",
    checksum: "d3fc8abc2d628307f78dd563211c37d7e87c7c90161dead6317d2367df4ba90f"
)

v0.0.7

SpeechCore XCFramework

Checksum: 704ed1d7586fbc236ad6916cb923dafb04068edab57d328ceffcbfe152a10c49

Usage in Package.swift

.binaryTarget(
    name: "CSpeechCore",
    url: "https://github.com/soniqo/speech-core/releases/download/v0.0.7/SpeechCore.xcframework.zip",
    checksum: "704ed1d7586fbc236ad6916cb923dafb04068edab57d328ceffcbfe152a10c49"
)

v0.0.6

SpeechCore XCFramework

Checksum: aca6733cd04b873e1f7a428993e8d4f23ffceed42f7507cd1196c0b89d34f170

Usage in Package.swift

.binaryTarget(
    name: "CSpeechCore",
    url: "https://github.com/soniqo/speech-core/releases/download/v0.0.6/SpeechCore.xcframework.zip",
    checksum: "aca6733cd04b873e1f7a428993e8d4f23ffceed42f7507cd1196c0b89d34f170"
)

v0.0.4

SpeechCore XCFramework

Checksum: 5e48d536134d147ac0162c67df39d59a60bdfc0f9c2793e81e4bb6046f06d15f

Usage in Package.swift

.binaryTarget(
    name: "CSpeechCore",
    url: "https://github.com/soniqo/speech-core/releases/download/v0.0.4/SpeechCore.xcframework.zip",
    checksum: "5e48d536134d147ac0162c67df39d59a60bdfc0f9c2793e81e4bb6046f06d15f"
)

v0.0.5

SpeechCore XCFramework

Checksum: ec87ae9191875390e19cd2aa74bfdbd8f314c4a0e83dfe012eed4d1ca30c4a5d

Usage in Package.swift

.binaryTarget(
    name: "CSpeechCore",
    url: "https://github.com/soniqo/speech-core/releases/download/v0.0.5/SpeechCore.xcframework.zip",
    checksum: "ec87ae9191875390e19cd2aa74bfdbd8f314c4a0e83dfe012eed4d1ca30c4a5d"
)

v0.0.3

SpeechCore XCFramework

Checksum: f63ea912c8c8b556d16aa197410ff33ee839915806a5ce31a9b8bc0e663eff04

Usage in Package.swift

.binaryTarget(
    name: "CSpeechCore",
    url: "https://github.com/soniqo/speech-core/releases/download/v0.0.3/SpeechCore.xcframework.zip",
    checksum: "f63ea912c8c8b556d16aa197410ff33ee839915806a5ce31a9b8bc0e663eff04"
)

v0.0.2

SpeechCore XCFramework

Checksum: 67a714baf4ada80dd550a40da0c31f1dcd2be1fd41f2cbf3b45bf0891d8afbc7

Usage in Package.swift

.binaryTarget(
    name: "CSpeechCore",
    url: "https://github.com/soniqo/speech-core/releases/download/v0.0.2/SpeechCore.xcframework.zip",
    checksum: "67a714baf4ada80dd550a40da0c31f1dcd2be1fd41f2cbf3b45bf0891d8afbc7"
)

v0.0.1

SpeechCore XCFramework

Checksum: 23ae4a62051b8eec3710a4794ac26ff202c4bdec746957c51ca89db41c39dc0f

Usage in Package.swift

.binaryTarget(
    name: "CSpeechCore",
    url: "https://github.com/soniqo/speech-core/releases/download/v0.0.1/SpeechCore.xcframework.zip",
    checksum: "23ae4a62051b8eec3710a4794ac26ff202c4bdec746957c51ca89db41c39dc0f"
)