LinguaMic LogoLinguaMic
Documentation

Open Source Licenses

LinguaMic is built on the work of the open source community. We acknowledge and thank the developers and communities behind these projects.

TermsPrivacyAboutAttributions
01

What We Use

LinguaMic's voice pipeline is an integration of several open-source projects. We have not modified the underlying model weights of any of these projects. Our customizations sit in the prompt-formatting, chunking, tone-preset, and serving layers we wrote on top.

Orpheus TTS

Developer: Canopy Labs

License

Apache License 2.0

Usage

The speech-synthesis model that powers our voice generation

Repository

github.com/canopyai/Orpheus-TTS

orpheus-tts-local

Developer: Isaiah Bjork

License

Apache License 2.0

Usage

Local-inference client we use to run Orpheus through LM Studio, including the GGUF model quantization

Repository

github.com/isaiahbjork/orpheus-tts-local

SNAC

Developer: Hubert Siuzdak

License

MIT License

Usage

Neural audio codec that decodes Orpheus's token output into 24 kHz waveforms

Repository

github.com/hubertsiuzdak/snac

faster-whisper

Developer: Guillaume Klein (SYSTRAN)

License

MIT License

Usage

High-performance CTranslate2 implementation of OpenAI Whisper, powering our speech-to-text endpoint

Repository

github.com/SYSTRAN/faster-whisper

Whisper

Developer: OpenAI

License

MIT License

Usage

The underlying speech-recognition model architecture and weights used by faster-whisper

Repository

github.com/openai/whisper
LinguaMic is not affiliated with or endorsed by Canopy Labs, Isaiah Bjork, OpenAI, SYSTRAN, or the SNAC author. All trademarks are the property of their respective owners.
02

Our Customizations

We have not retrained or fine-tuned the Orpheus speech model. The underlying weights are the unmodified canopylabs/orpheus-3b-0.1-ft checkpoint distributed by Canopy Labs.

What we do build on top:

  • A FastAPI service that wraps the inference pipeline and exposes our internal API.
  • A corrected prompt-formatting layer that adds Canopy's audio-primer tokens that the upstream GGUF client omits.
  • A library of tone presets (calm, romantic, storytelling, horror, angry, adventurous, excited, sad, funny) that drive Orpheus's sampling parameters.
  • Text sanitization that strips emotion tags not in the Orpheus vocabulary and splits CamelCase brand names so the model pronounces them correctly.
  • A chunking layer that splits long input at sentence and clause boundaries while keeping emotion tags inline.

The Apache License 2.0 full text is available at apache.org/licenses/LICENSE-2.0. The MIT License full text is available at opensource.org/licenses/MIT.

03

Apache License 2.0 - Summary

The Apache License 2.0 is a permissive free software license that allows users to use, modify, and distribute the software for any purpose, subject to the terms and conditions of the license.

  • OKCommercial use permitted
  • OKModification permitted
  • OKDistribution permitted
  • OKPatent use permitted
04

Contact

Questions about our use of open source software? Contact us.