Changelog

Added optional InflectSynth synthesis via the canonical inflect engine ID

Added optional AudioExport profile exports for project audio and M4B

Changed profile mode to leave legacy exports and speak streaming unchanged

BREAKING: Removed Readio-owned SSMD authoring and template workflows

Added Kitten engine discovery, voice selection, diagnostics, and synthesis through KittenSynth

Added ssmdconvert source and converter provenance to auto-converted documents

Added SupertonicSynth support and managed PocketSynth voice prompts

BREAKING: Changed configuration and project roles to engine-qualified schema-3 targets with explicit v0.3 migration

BREAKING: Changed engine IDs to kokoro, piper, pocket, supertonic, and kitten; removed provider selectors, voice aliases, and the v1 writer

Changed interactive playback to stream through Readio’s bounded queue while keeping mastering for file renders

Changed engine extra floors to released packages compatible with OnnxVoice 0.2

Documented canonical engines, explicit v0.3 migration, streaming playback, and conversion provenance

Added use_saved_settings for fresh projects.resolve_synthesis() candidates

Changed voice language filtering to honor generic capabilities and conflicting locales

Changed Piper and Pocket gender projection to preserve explicit metadata without inferring missing values

Changed voice metadata projection to preserve stable selectors independently of descriptive locales

Added mastering profiles with LUFS and true-peak reporting across composition workflows

Added saved project settings for synthesis, composition, export, and audiobook workflows with stage-aware rebuilds

Documented canonical SSMD syntax, prosody values, and .ssmd.md authoring guidance

Documented an audio-first evaluation corpus and rubric for standalone SSMD guides

Added formatted command help and engine and output-format discovery to the CLI

Added lexicon, pronunciation, and language-detection policies to the read-only synthesis preflight API

Changed the default short-sentence policy to phrase, with migration for legacy auto settings

Changed project status output with clear issue explanations and actionable commands in the CLI and public API

Documented the v0.3.1 public API addition in the generated changelog

Added audiobook chapter descriptions and non-rendering synthesis resolution to the public API

Changed Readio’s runtime integration for fixed timestamp-capable releases and OnnxVoice diagnostics

Requires OnnxVoice 0.1.8+ and exposes additive stable en-US Kokoro selectors through voices list –engine kokoro –lang en-us without changing de-ko-* identities

Added observable composition progress for segment processing, operations, assembly, loudness, and artifact writing

Documented compose progress flags, stderr separation, approximate ETA scope, and final composition phases

Added explicit spaCy model tiers and configurable short-sentence synthesis policies

Added append-only OnnxVoice identities with engine-qualified selectors for Kokoro and Piper

Added backend-neutral lexicon discovery and CLI inspection with language, model, and engine filters

Fixed lexicon listing without a language filter while preserving backend-aware selector disambiguation

Changed Readio dependency metadata and CI to use PyKokoro 0.9.5 and SSMD 0.8.7+

Changed package publishing with artifact checks and release-tag version validation

Changed spaCy model-size and short-sentence synthesis controls, including reader configuration migration

Changed Readio’s speak/render pause default from tts to auto while preserving config and CLI overrides

Documented the PyKokoro support floor, 0.9.5 tested target, and supported SSMD API

Improved PyKokoro compatibility and package metadata regression coverage

Added synthesis plan resolution system with explicit plan domain, resolution pipeline, and validation

Added manifest module for project metadata and package information management

Improved plan feature with enhanced CLI commands and validation

Added comprehensive LLM guides for SSMD audio content creation including podcasts, stories, educational content, and more

Added positional file detection for speak, render, and Spotify publish with clear literal-text and missing-path handling

Added model-specific voice and quality discovery, validated language defaults, ordered lexicons, and frontend checks

Changed PyKokoro integration to the 0.9 public model registry and runtime contract

Improved offline registry, runtime asset, and PyKokoro API/version diagnostics

Changed model-source selection and model-scoped voice resolution to apply consistently across synthesis and SSMD

Added explicit Spotify publishing, direct uploads, readiness inspection, and chapter timeline controls

Added runtime model discovery and validated per-language synthesis defaults with CLI overrides

Added SSMD voice binding workflows for reusable roles and one-off render resolution

Added MP3, M4A, and OGG output for rendering and Spotify publishing

Added terminal progress reporting for rendering and Spotify publishing workflows

Fixed package installation by declaring the typing_extensions runtime dependency

Added speech-friendly Markdown input for files, stdin, and literal text

Added local text playback from literals, files, standard input, and live paragraph streams

Added bounded-memory WAV rendering and explicit Spotify publishing with readiness and chapter support

Added ingest artifacts, templates, SSMD voice preflight, configuration validation, and actionable diagnostics

Added cross-platform CI, pre-commit, coverage, package publishing checks, and a podcast example