Editor & Export

Editor & export

The Editor tab shows your audiobook line by line ("chunks"), with tools to edit, re-render, listen and check each one. The Result tab exports the finished book.

The chunk editor

Each chunk shows:

• Status — Pending (no audio yet), Done, or Error;
• Speaker — editable;
• Text — the spoken line or narration, editable;
• Instruct — the delivery instruction, editable.

Each chunk can also be played, re-rendered on its own, and inserted or deleted.

Editing a line

1. Click a field and edit it; changes save automatically.
2. Click the chunk's generate button to re-render just that line — useful for fixing a line after an edit, trying a different instruction, or comparing voices.
3. Or click Render Pending to render every chunk that still needs audio.

Rendering

• Render Pending renders every chunk that has no audio yet.
• Regenerate All re-renders every chunk.
• With the built-in TTS (local mode) both use batched rendering, about 3–6× real time. With an external TTS server they send one request per line in parallel. See Batch generation.
• Both respect the GPU lock: they will not start while script generation is using the card, unless the language model is marked as not on this GPU.
• Cancel stops a render; finished chunks are kept.

Checking the audio

• Check Voices compares each rendered chunk with its speaker's reference voice (ECAPA similarity) and flags chunks that drifted. Tick flagged only to see just those.
• Text integrity shows, word by word, where the script differs from the source text.
• Every generated file is validated as real audio; a broken file is flagged on its chunk.

Listening

• Play a single chunk with its play button.

• Play Sequence plays chunks in order, with natural pauses:

  • between different speakers — 500 ms by default (Speaker Change Pause);
  • between lines from the same speaker — 250 ms by default (Same Speaker Pause).

  Both are set in the Setup tab.

Merging

Click Merge All to join all chunks that have valid audio into one audiobook, in order, with the pauses above. Silence at the start and end of each line (dead air — on LoRA voices a median of 310–340 ms) is trimmed when the chunks are joined, so the pauses you set are the pauses you hear. The result appears in the Result tab.

Export options

The combined audiobook

• One MP3 at 128 kbps with every voice and the natural pauses.
• Download it from the Result tab; the file is cloned_audiobook.mp3.

M4B with chapters

Export M4B makes an audiobook file with chapter markers. You can set the Title, Author, Narrator, Year, Description and Cover Image, and choose Per-chunk chapters.

Chapter files

Export chapters writes each chapter as its own MP3 or WAV file, named by a template such as {chapter_number} - {chapter_name}. Options include number padding, book and series names, volume, a chapter list, per-chunk files, changed chapters only, and a preview of the file names. The files and a zip of them are in chapter_exports/.

Individual lines

Each line is also saved as its own file in voicelines/, numbered in reading order and named by speaker, ready to import into audio software:

voicelines/
├── voiceline_0001_narrator.mp3
├── voiceline_0002_elena.mp3
├── voiceline_0003_marcus.mp3
└── ...

Audacity

Export to Audacity (Result tab) makes one WAV track per speaker for multi-track editing:

audacity_export.zip
├── project.lof       # open in Audacity to import every track at once
├── labels.txt        # import via File > Import > Labels for line annotations
├── narrator.wav      # full-length track with only the NARRATOR's audio
├── elena.wav         # full-length track with only ELENA's audio
├── marcus.wav        # full-length track with only MARCUS's audio
└── ...

To use it:

1. Download and unzip audacity_export.zip.
2. Open project.lof in Audacity; every speaker's track is imported.
3. Import labels.txt via File > Import > Labels for the line annotations.

Every track is padded with silence to the same length, so playing them all together sounds exactly like the merged MP3. This is useful for per-character volume, effects on one speaker, adjusting the timing between lines, and professional post-production.

Broken or tiny MP3 files

A 428-byte MP3 means ffmpeg is missing or cannot encode MP3. The installer builds ffmpeg into the app's environment, so this usually means the app is running outside app/env. Start it from Pinokio (or activate app/env) and render again. See Troubleshooting.