Smart Detect¶

The Smart Detect workflow finds chapter cues by analyzing the audio itself. This is the recommended workflow when your book has no usable Sources.

When to use it¶

• New book with no chapters.
• Existing chapters are incorrect.
• You are not sure what you have and want to see what Achew finds.

Steps¶

1. Select a book from Achew's main screen.
2. Choose Smart Detect on the Select a Workflow screen.
3. Toggle the Dramatized checkbox for books that contain music or sound effects. Note that dramatized detection takes significantly longer than regular detection, so it is recommended to leave this unchecked for regular audiobooks.
4. (Optional) Add or inspect existing Chapter Sources from this screen. These can be used as a comparison reference once detection finishes.
5. Click Start Smart Detect, and Achew will begin analyzing the audio. This process can take anywhere from a few seconds to several minutes depending on the length of the book, how powerful your machine is, and whether Dramatized was checked.
6. After detection has finished, use the Initial Chapter Selection screen (see below) to decide which cues will become chapters, then click Create Chapters.
7. On the Transcribe Titles screen, choose your transcription settings and click Transcribe. The first few seconds of audio at each chapter timestamp will be transcribed into a title.
8. After transcription has finished, use the Chapter Editor to review and polish the chapters.

Initial Chapter Selection¶

Once audio analysis completes, you'll be taken to the Initial Chapter Selection screen. Here you will choose which detected cues to convert into chapters before heading into the Chapter Editor.

Audiobooks differ wildly in pacing, narration style, and mastering. There is no single "correct" setting, so Achew gives you interactive tools to find the right balance for your book.

Timeline Visualization¶

The timeline visualizes how the selected cues are distributed throughout the audiobook. Each tick mark represents one selected cue (i.e. a chapter that will be created). Use the tools explained below to adjust which cues are selected. You'll typically want to aim for a semi-even distribution of cues across the book's length.

The Cue Selection Slider¶

Use the main slider to increase or decrease the number of cues that will be turned into chapters.

The bar chart behind the slider displays a histogram showing the distribution of the silence gaps that precede the detected cues:

• Cues on the left have the longest silence gaps and are highly likely to represent real chapters.
• Cues on the right have shorter gaps and are more likely to be false positives (like long pauses between sentences or paragraphs).

Moving the slider to the left will select fewer cues, while moving it to the right will sweep in more cues with shorter gaps. Adjust the slider until the chapter count and timeline distribution roughly match what you expect for the book.

Later on you'll be able to both add and delete chapters, but deleting extraneous chapters is much easier than finding missing chapters, so at this stage it is recommended to err on the side of selecting more cues rather than fewer.

Comparing Alignment¶

If an existing source (e.g. embedded metadata or Audnexus), has accurate chapter timing, you can compare it to your selected cues by toggling that source in the "Compare to" section of the timeline.

Aim for a high alignment, indicated by the color of the tick marks and the alignment percentage.

Intro/Outro Sensitivity¶

The Intro/Outro Sensitivity slider adjusts how likely it is that cues near the beginning and end of the audiobook will be included in your selection. If you find that intro and outro tracks are being missed (prologues, epilogues, publisher dedications, etc.), use a higher sensitivity. Conversely, you can lower the sensitivity if you prefer to exclude such tracks.

Including Unaligned Chapters¶

If there are one or more Chapter Sources, you'll see a section titled Include unaligned chapters near the bottom of the page, with a checkbox for each source. The count in parentheses next to each source is the number of source chapters that don't line up with any of the selected cues.

Checking the box for a source will cause its timestamps to be included when chapters are created, on top of the cues already selected by the slider. This is helpful when a source has accurate chapter breaks that Smart Detect couldn't find on its own.