v0.5.0

What's Changed

• Add separate narrator and dialogue voice paths while keeping voice_path as the shared fallback.
• Work around Chatterbox token-repetition EOS false positives and clear leaked attention hooks between generations.
• Increase the default comma, semicolon, and colon pause to 200ms.

Verification

• uv run pytest -q
• uv build

v0.4.3

Changes

• Preserve real silence pauses during segment compaction.
• Only compact adjacent same-kind segments when no existing pause would be lost.
• Keep sentence, paragraph, and dialogue transition pauses as actual generated silence.
• Add tests for pause-preserving compaction.

Verification

• uv run pytest
• uv build --out-dir /tmp/chatterbook-release-0.4.3

v0.4.2

Changes

• Compact adjacent same-kind audio segments up to max_chars to reduce TTS calls.
• Preserve narration/dialogue boundaries during compaction.
• Keep batch groups separated by segment kind so narration and dialogue settings are not mixed.
• Add tests for dialogue boundary handling and same-kind compaction.

Verification

• uv run pytest
• uv build --out-dir /tmp/chatterbook-release-0.4.2

v0.4.1

Changes

• Add batch_size to Book.convert().
• Group adjacent narration/dialogue segments into same-style batches.
• Prepare voice conditionals once per batch when voice_path is used, avoiding per-segment voice prompt preparation.
• Document batch_size and add tests for batch grouping and conditional reuse.

Verification

• uv run pytest
• uv build --out-dir /tmp/chatterbook-release-0.4.1

v0.4.0

Changes

• Introduce the public Book API for serialized EPUB analysis and conversion.
• Replace convert_epub with Book(...), to_dict(), from_dict(), and convert().
• Store narration/dialogue segments and pause timing in the serializable book representation.
• Improve conversion progress display and reuse the serialized audio plan during TTS rendering.

Verification

• uv run pytest
• uv build --out-dir /tmp/chatterbook-release-0.4.0

v0.3.0

Release v0.3.0\n\n- Preserve EPUB paragraph boundaries instead of merging paragraphs together.\n- Add natural audiobook pacing defaults: speed=0.9, paragraph pauses, and dialogue pauses.\n- Split quoted dialogue into separate expressive audio segments.\n- Add dialogue_exaggeration and dialogue_cfg_weight controls.\n- Adjust M4B chapter metadata for tempo changes.\n- Raise GenerationError with chapter title and text preview when TTS generation fails.

v0.2.3

Release v0.2.3\n\n- Suppress Chatterbox alignment/repetition warning logs during conversions.\n- Keep the internal repetition guard active while hiding noisy implementation details from users.

v0.2.2

Release v0.2.2\n\n- Hide Chatterbox internal Sampling progress output during conversions.\n- Keep the single green EPUB progress bar as the only visible progress indicator.\n- Add tests for model progress suppression.

v0.2.1

Release v0.2.1\n\n- Show one colored tqdm progress bar per EPUB instead of nested paragraph/chunk bars.\n- Add a pkg_resources compatibility shim so Chatterbox/resemble-perth works with setuptools >=82.\n- Keep M4B output behavior from v0.2.0 unchanged.

v0.2.0

Release v0.2.0\n\n- Make M4B the default audiobook output.\n- Use EPUB title metadata for the default output filename.\n- Embed chapter metadata into generated M4B files with ffmpeg.\n- Keep chapter WAV export via output_format="wav".\n- Add colored tqdm progress bars for chapter and chunk processing.\n- Group EPUB paragraph blocks for more stable Chatterbox generation.