Skip to content
loudkit

What loudkit 0.1 supports

The scope, stated once, so the README cannot promise more than the project keeps. Anything not listed under Supported is either Experimental (works, shipped, feedback wanted, may change) or Out of scope for 0.1 (not attempted). Removing something from Supported is a breaking change. Promoting something into it is not.

Supported

Section titled “Supported”
  • Local synthesis, offline. No account, no network dependency, no per-character billing. The engine loads a packed checkpoint and runs on your hardware.
  • Voice profiles as files. A voice is a ~150 KB .safetensors you can copy, mail and keep; cloning one takes about ten seconds of audio, from the shell with loudkit clone or from any SDK with enroll.
  • Deterministic rendering. Same text, voice, seed and build give a bit-identical waveform on a given device and backend, with the contract and its limits written down in docs/reference/IDENTITY-CONTRACT.md.
  • Python as the reference implementation, with the full API: synthesize, synthesize_long, streaming with token-level cancellation, previous_tokens continuation, speed control, provenance-carrying WAVs.
  • Backends: torch (CPU, CUDA, MPS) as the reference execution, ONNX Runtime as the portable no-torch deployment, CoreML for Apple. Execution may differ in speed per backend, never in what is computed.
  • Compatibility ports: Swift, Go, Rust and TypeScript implement the same algorithm against shared conformance fixtures. They track the reference: features land in Python first and reach the ports with the fixture that proves them.
  • The CLI, eight commands: speak, clone, voices, download, serve, verify, doctor, grpc. loudkit --help lists these and no others. describe, bench and profile stay in the package as repo tools and stay runnable, and mcp is preview; none of the four is part of this contract.
  • Transports over one synthesis path: the local HTTP server (/v1, SSE streaming, OpenAI-compatible route), gRPC (loudkit grpc, typed schema and streaming backpressure, contract in proto/loudkit.proto), and Speech Dispatcher. All answer byte-for-byte what the library answers, with one frozen error-code catalog (docs/reference/errors.md).
  • Speech Dispatcher integration: loudkit as a system voice for Linux screen readers, with the protocol codes and rate control handled properly.
  • Text handling in 12 languages (numbers, dates, currency, units, abbreviations), verbalised from first principles and fuzz-tested.
  • Voices for ten languages (en, pl, de, fr, nl, es, it, pt, sv, da), each with its donor or source, licence and consent basis recorded in docs/voices/roster/provenance.json.
  • Provenance: every saved WAV and server reply carries a C2PA claim-only manifest naming the algorithm, checkpoint, voice profile, backend and seed.

Supported with stated limits

Section titled “Supported with stated limits”
  • Language quality. We evaluated English by ear. We do not speak the other nine shipped languages well enough to judge their naturalness reliably. Automated checks cover them, but those checks are not a native-speaker verdict. Feedback from native speakers is very welcome.
  • Waveforms are not identical across devices or backends. Equivalence is measured and banded. Identity holds per build/device/backend.
  • --cuda-graphs is an opt-in throughput mode in the identity contract’s “equivalent” class: deterministic, not token-identical to eager.
  • ChunkConfig.first_chunk_max_tokens is Python-only for now. It is an algorithm-level knob that re-fingerprints when set. The ports follow the usual fixture-first route.
  • loudkit clone reads a local WAV or FLAC file. No URLs, no recording, no denoising, no trimming, no batch, no preview. --name and --language are stated, not guessed. The API path, loudkit.enroll(...), exists in every SDK’s language (Python, Swift, Go, Rust, TypeScript) and takes samples directly. There is no MCP clone tool.

Experimental

Section titled “Experimental”
  • MCP over stdio (loudkit mcp). It answers the same bytes as every other transport and stays in the package, but it is not in the top-level help and its tool surface may change shape.
  • The systemd unit and other deployment scaffolding in integrations/ beyond the Speech Dispatcher module.
  • The OpenAI-compatible route’s parameter mapping outside the documented fields.

Out of scope for 0.1

Section titled “Out of scope for 0.1”
  • Equal listening quality in every language.
  • A production public server. loudkit serve is a local runtime with a hardened loopback default, not a multi-tenant service: no TLS, no quotas, no per-voice authorization.
  • Expressiveness or emotion control.
  • Robotics integrations (ROS, Wyoming), Android, WebAssembly, additional backends beyond the three shipped.
  • Model training and fine-tuning.