Audio pages

Audio pages live in the book like any other page. Use them to add an oral history, an interview, or a recording of someone reading a letter aloud. They show up in the reading order, get transcripts, and let you mark who each speaker is so chat and search know who’s talking.

Audio pages vs. audio recordings

Two ways to add audio to the book:

• Audio pages — first-class pages in the book, appear in the reading order, get a transcript, and support speaker review. Use these for longer recordings: stories, oral histories, interviews.
• Voice memo recordings — short notes attached to an existing image or item. Don’t change the page count. See the Audio recordings page.

Adding an audio page

1. Open the book and switch to expert view.
2. Use the admin Add pages menu and choose Audio file.
3. Pick the audio file from your computer. Supported formats: MP3, M4A, WAV, FLAC, OGG, WebM.
4. The browser uploads the file directly to storage and the new page appears at the end of the book (or at the position you chose).

Files up to 100 MB are supported. The upload runs in the background — you don’t need to keep the page open.

What the audio page looks like

Readers see a native audio player and a transcript. If you upload a display photo, that photo appears next to the player. Otherwise a spoken-word placeholder — person silhouette and soundwaves — appears so readers know it’s a recording.

Audio pages keep the usual page affordances: captions, comments, reordering, and deletion. Image-only actions — cropping, rotating, photo enhancement, page-item selection — are hidden because they don’t apply.

Setting or replacing the display photo

1. Open the audio page.
2. From the page action menu (⋮), choose Set display photo.
3. Pick a JPG, PNG, WebP, or AVIF up to 10 MB.
4. The photo replaces the placeholder right away in the viewer and the thumbnail strip.

Display photos are a passive surface — they aren’t OCR’d, face-detected, or enhanced. You can replace or remove the photo at any time from the same menu.

Transcripts

When the family has granted the OpenAI processor consent, transcription runs automatically after upload. Status messages appear in the transcript panel:

• Transcript: generating… — transcription in progress.
• Transcript failed — an error occurred; an admin can retry from the page action menu.
• Otherwise the transcript appears under the audio controls.

When the configured model supports speaker diarization, the transcript renders as turn-by-turn paragraphs labeled Speaker A: …, Speaker B: …, and so on. Plain transcripts (no speaker labels) render as a single block of text.

Speaker review

For diarized transcripts a Who is each speaker? section appears under the transcript. Each unique speaker label gets a row:

• Link… opens an inline form. Type a person’s name and press Link existing to connect this speaker to someone already in the family directory.
• Create new in the same form makes a new person record and links the speaker in one step. Use this for someone who doesn’t have a page or face yet.
• Skip marks the speaker as dismissed for now — useful when a voice in the background isn’t the focus.
• Unlink on a linked speaker reverts to unresolved if you change your mind.

Once a speaker is linked, the transcript shows the person’s name in place of the raw label, and the person is added to the page’s list of people. Chat, search, and the knowledge graph then treat the speaker’s turns as that person’s words.

Speaker review is per-page in v1 — linking the same voice in a different recording is a separate action. Cross-recording voice matching will come later.

Deleting an audio page

Deletion behaves like any other page: select Delete page from the page action menu. The audio file, display photo, transcript, and speaker review state are all removed.