Portfolio

Open Source

Our open-source work — FOSS, self-hostable libraries and tools spanning voice AI, data extraction, datasets, NLP, and games, plus the OpenVoiceOS and HiveMind ecosystem we help build. No lock-in, no mandatory cloud.

Beyond the code, our open datasets and models are published onHugging Face,and our training notebooks live in theml-notebooksrepository.

Community

Working with the community

We contribute to independent partners and donate publicly-funded work to the OpenVoiceOS foundation.

Alpha Cephei

A speech-recognition partnership bringing decades of automatic speech recognition (ASR) expertise — the team behind Vosk and Kaldi-based models — to our voice-model fine-tuning and dataset work.

ILENIA

Spanish program advancing language technology across the nation's official languages (Spanish, Catalan, Basque, Galician). Funding OVOS development to create downloadable, privacy-respecting voice assistants that run on Raspberry Pi and personal computers. Funded by Spain's Ministry of Digital Transformation and EU Recovery Plan.

NGI0 Commons Fund / NLnet

European Commission's Next Generation Internet program (with Swiss SERI co-funding) supporting OpenVoiceOS as a stable, privacy-first alternative to Big Tech voice assistants. Focus: multi-language support, user onboarding, platform stabilization, and documentation.

OpenVoiceOS

An independent non-profit foundation building a community-driven, privacy-respecting open-source voice-AI platform. TigreGótico is a core contributor to OpenVoiceOS, and its founder serves on the foundation's board.

Explore

Featured highlights

Filter by area to explore more — everything is open source.

Developer Tools

agentpipe

Async Python wrapper for coding-agent CLIs (Claude Code, Gemini, opencode, Antigravity) behind one Provider API — one-shot, multi-turn sessions, typed streaming events, and fan-out/delegate/retry pipeline patterns.

  • agents
  • async
  • CLI
  • orchestration
Self-hostedView repository
Phonetics & G2P

ahotts-g2p

Pure-Python, zero-dependency, version-aware grapheme-to-phoneme for Basque and Spanish — faithfully reproducing the AhoTTS front-end and emitting the single-char IPA string used to train StyleTTS2/VITS voices.

  • G2P
  • Basque
  • Spanish
  • IPA
Self-hostedView repository
Speech Recognition

ALIGN

A browser-based tool for aligning audio with text transcriptions and IPA at word, grapheme, and sentence level. Runs entirely client-side, no server, no dependencies, no build step.

  • IPA
  • audio-text alignment
  • browser-based
  • JavaScript
Self-hostedView repository
Data & Scraping

anon_requests

Anonymous HTTP with IP rotation; composes with unblock_requests for rotation plus bot-wall bypass.

  • HTTP
  • IP rotation
  • privacy
Self-hostedView repository
Phonetics & G2P

arbtok

Rule-based Arabic (MSA) text-to-IPA with tashkeel diacritization, built on orthography2ipa — a context-sensitive token tree handles sun-letter assimilation, hamzat al-waṣl, tanwīn and tāʾ marbūṭa, diacritizing bare text first via bundled ONNX models.

  • Arabic
  • G2P
  • IPA
  • diacritization
Self-hostedView repository
Media & Music

audiobooker

Search and stream public-domain audiobooks from LibriVox and similar sources.

  • LibriVox
  • audiobooks
  • scraper
Self-hostedView repository
Speech Synthesis

audiosronnx

Speech restoration, denoising and bandwidth extension in pure ONNX — upscales narrowband speech to full 48 kHz without torch at runtime.

  • ONNX
  • audio super-resolution
  • speech enhancement
Self-hostedView repository
Developer Tools

baresipy

Minimal HTTP client library for Python — lightweight alternative to requests.

  • HTTP
  • Python
  • client
  • lightweight
Self-hostedView repository
Phonetics & G2P

bifonia

European-Portuguese heterophonic homograph disambiguation for TTS — picks the right pronunciation of words spelled alike but read differently by meaning (sede = thirst vs headquarters, forma = mould vs shape). Pure-Python rules plus learned models, with open datasets.

  • grapheme-to-phoneme
  • TTS
  • homographs
  • word-sense disambiguation
Self-hostedView repository
NLP & Language

brill_postaggers

Pre-trained Brill part-of-speech taggers for 11 languages, shipped as ready-to-use pickled models over NLTK — tag text without training.

  • POS tagging
  • 11 languages
  • NLTK
Self-hostedView repository
Media & Music

classical-archives

Python package to fetch information about classical composers from Classical Archives.

  • Classical music
  • API client
  • Music archives
Self-hostedView repository
NLP & Language

crf_query_xtract

Multilingual CRF keyword/search-term extractor — pulls the query out of a natural-language question using a CRF over POS and orthographic features, with pretrained per-language models.

  • keyword extraction
  • CRF
  • multilingual
Self-hostedView repository
Phonetics & G2P

desacordo_ortografico

Detect and convert between seven Portuguese orthographic norms — the pre-1911 etymological writing, the 1911/1943/1945/1971/1973 reforms, and the Acordo Ortográfico de 1990 with its European and Brazilian sub-variants.

  • Portuguese
  • orthography
  • AO1990
  • text conversion
Self-hostedView repository
Phonetics & G2P

espyak

Pure-Python, dependency-free reimplementation of espeak-ng's grapheme-to-phoneme front-end — text to phonemes only, no C extension. Reproduces the espeak-ng binary byte-for-byte across an 86-language headword sweep and a 31-language sentence corpus; 117 languages bundled.

  • G2P
  • IPA
  • phonemization
  • espeak-ng
Self-hostedView repository
Phonetics & G2P

euskaphone

Dialect-aware Basque text-to-IPA phonemizer built on the shared orthography2ipa pronunciation lattice — open, inspectable and source-cited.

  • Basque
  • G2P
  • IPA
  • TTS frontend
Self-hostedView repository
Voice PlatformsFeatured

HiveMind

A protocol for building hierarchical networks of agents — lightweight satellite devices connect over an encrypted mesh to any conversational agent, from OpenVoiceOS to LLM personas to arbitrary A2A agents. The ecosystem spans transport protocols, voice satellites, chat bridges, cryptography and zero-config discovery.

  • agent networks
  • distributed voice
  • protocol
Self-hostedView repository
NLnet

hivemind-a2a-agent-plugin

A2A agent protocol plugin for hivemind-core — bridges the hive to external Agent-to-Agent (A2A) servers over JSON-RPC 2.0, streaming responses back to satellites.

  • A2A
  • agent
  • interoperability
Self-hostedView repository
Smart Home

linux2mqtt

Publish a Linux/SBC's power and system telemetry to Home Assistant over MQTT — CPU/GPU/RAM/disk, temperatures, throttling, audio + MPRIS, WiFi/Bluetooth — all auto-discovered. Built on powerguess.

  • MQTT
  • Home Assistant
  • power
  • system monitoring
Self-hostedView repository
Media & Music

media-archivist

Index, canonicalize, deduplicate and serve media catalogues from YouTube and other sources.

  • catalogue
  • homelab
  • indexing
  • media
Self-hostedView repository
Media & MusicFeatured

mediavocab

A shared media-metadata vocabulary and normalization layer that unifies how titles, artists, and identifiers map across media clients and Music Assistant integrations.

  • media metadata
  • normalization
  • vocabulary
Self-hostedView repository
Media & Music

metadatarr

Pydantic-powered media-metadata clients and a keyless cross-source entity resolver.

  • Pydantic
  • media metadata
  • resolver
Self-hostedView repository
Developer Tools

nupyml

A scikit-learn-like machine-learning and neural-network library with only numpy and scipy as dependencies — written to be read.

  • machine learning
  • numpy
  • education
Self-hostedView repository
Voice PlatformsFeatured

Open Voice OS

Community-driven, privacy-first open-source voice assistant platform. TigreGótico is a core contributor to the whole ecosystem — maintaining speech plugins, STT/TTS services, developer tooling and formal spec work across dozens of repositories.

  • FOSS
  • privacy-first
  • voice platform
Self-hostedView repository
Developer Tools

OpenTone

Data-over-sound toolkit in pure Python — carry text and data through an audio channel (DTMF and more) using shared WAV/sine/Goertzel primitives. Core schemes are dependency-free.

  • data-over-sound
  • DTMF
  • audio
  • signal processing
Self-hostedView repository
Phonetics & G2PFeatured

orthography2ipa

Grapheme-to-IPA and allophone mappings for 350+ languages across 20+ families — a maximal-munch IPA tokenizer, phonological and script distance metrics, dialect transforms, and a pluggable G2P backend.

  • 350+ languages
  • G2P
  • IPA
  • phonemization
Self-hostedView repository
NLnet

ovos-busmon

Messagebus monitor for OpenVoiceOS — watch bus traffic live to debug skills, intents and services.

  • messagebus
  • debugging
  • monitoring
Self-hostedView repository
NLnet

ovos-localize

A GitHub-native localization platform for OpenVoiceOS skills and plugins — manage translations entirely through Git, with no external service to run.

  • GitHub-native
  • i18n
  • localization
  • no-infra
Self-hostedView repository
ILENIA

ovos-opendata-server

Open-data metrics server for OpenVoiceOS — collects anonymous usage metrics and publishes them as open datasets; developed under the ILENIA project.

  • metrics
  • open data
  • server
Self-hostedView repository
NLnet

ovos-plugin-arena

Benchmarking arena for OVOS plugins — head-to-head fights with ELO rankings to compare STT, TTS and other plugin implementations on real workloads.

  • benchmark
  • ELO
  • plugins
Self-hostedView repository
NLnet

ovos-stt-server

Host any OpenVoiceOS STT plugin as a network service — a small server exposing speech-to-text over HTTP for satellites and third-party clients.

  • STT
  • server
  • self-hosted
Self-hostedView repository
NLnet

ovos-tts-server

Host any OpenVoiceOS TTS plugin as a service — a simple flask server that turns any TTS plugin into a self-hosted speech-synthesis API.

  • TTS
  • server
  • self-hosted
Self-hostedView repository
NLnet

ovos_tts_transformer_FlashSR

Audio super-resolution for OVOS TTS — FlashSR upsamples synthesized speech from 16 kHz to 48 kHz via ONNX before playback, brightening any existing voice without retraining.

  • TTS
  • super-resolution
  • ONNX
Self-hostedView repository
NLnet

ovos_tts_transformer_NovaSR

Audio super-resolution for OVOS TTS — NovaSR upsamples synthesized speech from 16 kHz to 48 kHz with the fast FastSR upsampler, skipping audio already at 48 kHz.

  • TTS
  • super-resolution
  • ONNX
Self-hostedView repository
NLnet

ovoscope

End-to-end testing framework for OVOS skills with a lightweight in-process runtime.

  • OVOS skills
  • end-to-end testing
  • in-process runtime
Self-hostedView repository
NLnet, NLP & Language

palavreado

Dead-simple keyword-based intent parser — a drop-in Adapt replacement matching utterances against required/optional keyword slots; born at TigreGótico and donated to OpenVoiceOS.

  • intent parsing
  • NLP
Self-hostedView repository
Speech SynthesisFeatured

phoonnx

A Python library for multilingual phonemization and Text-to-Speech (TTS) using ONNX models.

  • ONNX models
  • multilingual TTS
  • phonemization
Self-hostedView repository
Speech Synthesis

phoonnx-AddonNVDA

Windows NVDA screen-reader add-on that speaks through phoonnx neural ONNX voices, demonstrating phoonnx as an accessibility TTS backend.

  • NVDA
  • accessibility
  • screen reader
  • ONNX
Self-hostedView repository
Speech Synthesis

phoonnx.js

In-browser VITS TTS inference with onnxruntime-web. Unicode and espeak-ng paths. No server, no API.

  • ONNX
  • TTS
  • browser
  • VITS
Self-hostedView repository
Data & Scraping

pombo_correio

Automated browser controller built on Selenium and Selenium Wire — sessions, event handling, element interaction, and extension management for scraping and testing where a full browser is unavoidable.

  • browser automation
  • Selenium
  • scraper
Self-hostedView repository
Smart Home

powerguess

Estimate or measure the power draw of a Linux device — a dependency-light Python library with provenance on every reading (INA219, RAPL, Raspberry Pi PMIC, battery, or a calibrated bounded estimate).

  • power
  • energy
  • RAPL
  • INA219
Self-hostedView repository
Wake Words

precise-onnx-js

Browser/Node.js port of precise-onnx wake word detection. MFCC feature extraction + ONNX-based detection compatible with Precise .onnx models.

  • ONNX
  • Wake words
  • Web
  • Audio
Self-hostedView repository
Smart Home

py-music-assistant

Python HTTP client and mediavocab bridge for a Music Assistant server — the shared transport and conversion layer behind the OVOS Music Assistant integration.

  • Music Assistant
  • media metadata
  • API client
Self-hostedView repository
ILENIA

pyAhoTTS

Python library for AhoTTS Text-to-Speech synthesis.

  • AhoTTS
  • ILENIA project
  • Python
  • TTS library
Self-hostedView repository
ILENIA

pyAhoTTS-Iparrahotsa

Python port of AhoTTS Iparrahotsa, a Text-to-Speech engine for Northern (continental) Basque — the Iparralde-dialect counterpart to pyAhoTTS, with hts_engine and the AhoCoder vocoder. Developed with Aholab (UPV/EHU).

  • AhoTTS
  • Basque
  • ILENIA
  • TTS library
Self-hostedView repository
Phonetics & G2P

pycotovia

Pure-Python port of the Cotovia G2P front-end for Galician and Spanish. Reproduces the compiled C binary's syllabification, stress and transcription pipeline with no subprocess.

  • G2P
  • Galician
  • Spanish
  • phonemization
Self-hostedView repository
Media & Music

pydiscogs

Streaming Python client for the Discogs monthly data dumps.

  • Discogs
  • Music metadata
  • Data dumps
  • API client
Self-hostedView repository
Data & Scraping

pyerowid

Typed Python client + markdown dataset dumper for Erowid (erowid.org).

  • Erowid
  • Harm reduction
  • API client
  • Dataset
Self-hostedView repository
Developer Tools

pyeye

Pure-Python port of the EYE N3 reasoning engine — feed it Notation3 facts and rules and it derives new facts via forward/backward chaining, RDFS and OWL 2 RL entailment, proof trees and 280+ built-ins. One dependency (rdflib).

  • N3
  • reasoner
  • RDF
  • forward chaining
Self-hostedView repository
Games

pyFrotz

Python wrapper for an interactive fiction engine.

  • Python
  • Z-machine
  • game
  • interactive fiction
Self-hostedView repository
Media & Music

pygutenberg

Python client for Project Gutenberg — book metadata via the open catalogue.

  • Project Gutenberg
  • Books
  • API client
  • Literature
Self-hostedView repository
Developer Tools

pyhermit

Pure-Python port of HermiT, the OWL 2 DL reasoner from Oxford — consistency, classification and instance retrieval via a hyperresolution tableau, with no JVM and zero runtime dependencies.

  • OWL 2
  • reasoner
  • symbolic
  • logic
Self-hostedView repository
Media & Music

pyimdb

Python metadata client for IMDb. Resolves free-text queries to canonical IMDb titles.

  • IMDb
  • Movies
  • Metadata
  • API client
Self-hostedView repository
Data & Scraping

pyinfopedia

Typed Python client for Infopédia, Porto Editora's European-Portuguese dictionary — parses each entry into IPA pronunciations, syllabification, etymology and senses, correctly separating heterophonic homographs for G2P and disambiguation work.

  • Portuguese
  • dictionary
  • IPA
  • lexicon
Self-hostedView repository
Data & Scraping

pyliveatc

Python client for LiveATC.net — discover, stream, and archive live and historical Air Traffic Control audio feeds, with dataset export for ASR training.

  • API client
  • ATC
  • audio
  • scraper
Self-hostedView repository
Media & Music

pymal

Scraper / API client for MyAnimeList — anime and manga.

  • anime
  • manga
  • scraper
Self-hostedView repository
Media & Music

pymetal

Encyclopaedia Metallum (Metal Archives) API client.

  • API client
  • Metal Archives
  • metal
Self-hostedView repository
Media & Music

pymusicbrainz

Python metadata client for MusicBrainz — the open music encyclopedia.

  • MusicBrainz
  • Music metadata
  • API client
Self-hostedView repository
Data & Scraping

pypsychonaut

Typed Python client + markdown dataset dumper for PsychonautWiki (psychonautwiki.org).

  • PsychonautWiki
  • Harm reduction
  • API client
  • Dataset
Self-hostedView repository
Data & Scraping

pyromhacking

Python scraper for romhacking.net (RHDN) — the community database of ROM hacks, fan translations, patching utilities, and documentation.

  • scraper
  • games
  • fan translation
Self-hostedView repository
Media & Music

pyshazam

A modular, async Python client for the Shazam API, built on top of shazamio_core for robust audio fingerprinting.

  • Shazam
  • Audio fingerprinting
  • Music recognition
  • API client
Self-hostedView repository
Data & Scraping

pysmwcentral

Python client for the smwcentral.net public JSON API — paginated access to the Super Mario World / SM64 / Yoshi's Island ROM-hacking catalogue.

  • API client
  • games
  • scraper
Self-hostedView repository
Data & Scraping

pytripsit

TripSit drug factsheets + interaction matrix client (harm-reduction).

  • TripSit
  • Harm reduction
  • Drug interactions
  • API client
Self-hostedView repository
Games

pyvndb

Python client for the VNDB (Visual Novel Database) kana public JSON API.

  • Visual Novels
  • Games
  • API client
  • VNDB
Self-hostedView repository
Data & Scraping

pywiktionary

Typed Python client for the English Wiktionary MediaWiki API.

  • Wiktionary
  • Dictionary
  • API client
  • Language
Self-hostedView repository
Data & Scraping

pywod

Multi-language Word-of-the-Day aggregator client.

  • Word of the Day
  • Language
  • API client
  • Dictionary
Self-hostedView repository
ILENIA

raspovos-audio-setup

Audio auto-configuration library for raspOVOS — detects sound cards, HATs and USB devices on Raspberry Pi and wires them up for voice assistant use; developed under the ILENIA project.

  • Raspberry Pi
  • audio
  • raspOVOS
Self-hostedView repository
Data & Scraping

remailers

Anonymous email remailer client with support for Cypherpunk and Mixmaster protocols.

  • anonymity
  • email
  • privacy
  • remailers
Self-hostedView repository
Phonetics & G2P

scriptconv

Zero-dependency script and phoneme-notation core — ISO-15924 detection, IPA ↔ ARPABET/X-SAMPA/Kirshenbaum/Cotovía, Buckwalter ↔ Arabic, Hangul → jamo, kana.

  • IPA
  • transliteration
  • phoneme notation
  • zero dependencies
Self-hostedView repository
Smart Home

shazam2mqtt

Dockerised bridge that fingerprints room audio with Shazam and publishes track metadata (title, artist, artwork, lyrics, Apple Music / Spotify / Deezer links) to MQTT with Home Assistant auto-discovery.

  • Shazam
  • MQTT
  • Home Assistant
  • music recognition
Self-hostedView repository
Data & Scraping

sitemapper

Site reconnaissance utility for learning site structure before building scrapers. Maps robots.txt, sitemaps, and link graphs using unblock_requests transport for resilient exploration of protected or complex sites.

  • web scraping
  • site mapping
  • robots.txt
  • sitemap
Self-hostedView repository
Speech Recognition

speakeronnx

Pure-onnxruntime speaker-embedding library — extract speaker embeddings, compute cosine similarity and verify speaker identity from ONNX models, with no PyTorch at runtime.

  • ONNX
  • speaker embeddings
  • speaker verification
Self-hostedView repository
Speech Synthesis, Speech Recognition

speechonnxmetrics

Unified speech evaluation metrics — neural MOS (UTMOS, NISQA, SIGMOS, DNSMOS), intrusive (STOI, SI-SDR, MCD) and ASR (WER/CER) — on numpy and onnxruntime only.

  • ONNX
  • evaluation
  • MOS
  • WER
Self-hostedView repository
Developer Tools

sshfs-keeper

A daemon that monitors SSHFS/rclone mounts and auto-remounts them when they drop — with a live web dashboard, REST API, Prometheus metrics, webhooks, and scheduled rsync/rclone sync jobs.

  • SSHFS
  • homelab
  • monitoring
  • self-hosted
Self-hostedView repository
Phonetics & G2P

stressonnx

Multi-language word-stress and accentuation placement for TTS front-ends, running on onnxruntime and numpy only.

  • ONNX
  • TTS frontend
  • word stress
Self-hostedView repository
Data & Scraping

temporalis

Unified multi-provider weather abstraction library for Python. Query OpenWeatherMap, Open-Meteo, MetNo, IPMA, NWS and more through one consistent API.

  • Weather
  • Multi-provider
  • API client
  • Open-Meteo
Self-hostedView repository
Phonetics & G2P

text2tashkeel

Lightweight Arabic diacritization (tashkeel) — a single tiny API over interchangeable ONNX models that restore the missing vowel marks, no PyTorch and offline by default. Powers arbtok's diacritization step.

  • Arabic
  • diacritization
  • tashkeel
  • ONNX
Self-hostedView repository
NLP & Language

tugalex

A lexicon handler and linguistic utility for Portuguese dialects.

  • NLP
  • Portuguese
  • lexicon
Self-hostedView repository
NLP & Language

tugamorph

Rule-based morphological analyzer for Portuguese — segments words into morphemes.

  • NLP
  • Portuguese
  • morphology
Self-hostedView repository
Phonetics & G2P

TugaPhone

Portuguese grapheme-to-phoneme tools.

  • Portuguese phonemes
  • dataset
  • phonetics
  • Portuguese
Self-hostedView repository
Media & Music

tunein

TuneIn radio scraper that emits typed media-metadata releases.

  • TuneIn
  • radio
  • scraper
Self-hostedView repository
Media & Music

tutubo

Lightweight YouTube scraper that emits typed media-metadata releases.

  • YouTube
  • scraper
  • video
Self-hostedView repository
Data & ScrapingFeatured

unblock_requests

A requests.Session subclass that bypasses Cloudflare (curl_cffi impersonation, FlareSolverr, Wayback fallback) — the anti-bot transport behind our scrapers.

  • HTTP
  • anti-bot
  • scraping
Self-hostedView repository
Data & Scraping

usenet

Live USENET reader and article scraper for alternative protocols research and archival.

  • USENET
  • privacy
  • protocols
  • scraping
Self-hostedView repository
Smart Home

vad2mqtt

Real-time Voice Activity Detection bridge using ovos-plugin-manager. Publishes speech probability (0–100 %), noise level (dB), and a debounced speech-detected binary sensor to MQTT — with Home Assistant auto-discovery.

  • VAD
  • MQTT
  • Home Assistant
  • OpenVoiceOS
Self-hostedView repository
Speech Recognition

vadonnx

Voice Activity Detection the ONNX way — load any VAD model behind one streaming API with a tiny runtime (numpy + onnxruntime). The VAD counterpart to phoonnx; most backends are data, not code.

  • ONNX
  • VAD
  • audio
  • streaming
Self-hostedView repository
Speech Synthesis

voiceclonnx

Pure-ONNX, multi-engine voice conversion — recast any speech to a reference speaker's voice with zero PyTorch at inference. 14 engines (kNN-VC, FreeVC, OpenVoice, RVC, CosyVoice…) behind one VoiceCloner API, plus a CLI and INT8 builds.

  • ONNX
  • voice conversion
  • zero-shot
  • multilingual
Self-hostedView repository
Wake Words

wakeforge

A wake-word model trainer developed under an NLnet grant for OpenVoiceOS — build custom, on-device wake words from your own samples.

  • NLnet
  • on-device
  • training
  • wake word
Self-hostedView repository
Wake Words

ww_tagger

Flask web app for hand-labeling wake-word audio — play each clip and tag word, speaker gender and noise type, persisting a corpus ready for training or evaluation.

  • labeling
  • dataset
  • Flask
Self-hostedView repository
Media & Music

xazam

Modular, async Python client for the Shazam API — audio fingerprinting via shazamio-core with typed Track/Artist models and a catalog scraper.

  • Shazam
  • async
  • audio fingerprinting
Self-hostedView repository
Developer Tools

z85base91

Compact binary-to-text encoding library (Base91, Z85B, Z85P) — fast C implementations with pure-Python fallbacks, all more compact than base64.

  • encoding
  • serialization
Self-hostedView repository