Speechify

Listen to your notes in a Speechify voice. The sentence being read is highlighted in the note and the word being spoken is highlighted inside it.

Press play and a small player appears over the bottom of the window with play and pause, sentence back and forward, the speed, and the voice. You can keep editing while it reads. Type above the playhead and the highlight moves with the text.

Install

The plugin is not in the community directory yet. Until it is, build it and copy three files into your vault:

git clone https://github.com/Speechify-AI/obsidian.git
cd obsidian
npm install
npm run build
mkdir -p <vault>/.obsidian/plugins/speechify
cp main.js manifest.json styles.css <vault>/.obsidian/plugins/speechify/

Then turn on Speechify under Settings, Community plugins. It needs Obsidian 1.11.4 or later.

Set up

You need a Speechify API key. Create one at https://platform.speechify.ai/api-keys, then open the plugin's settings and choose it under API key. Obsidian stores the key in its keychain. The plugin's own settings file, which lives in your vault and syncs with it, only holds the name you gave the secret.

The settings tab checks the key against the API as soon as you choose it and tells you how many voices it can use.

Use

• The Speechify icon in a note's header, or in the ribbon, reads the note from the top. With text selected it reads the selection.
• Right-click in the editor for Listen from here or Listen to selection.
• The player's voice button opens a searchable list of every voice your key can use, your own clones included, with a sample on each. Changing voice says the current sentence again in the new one.
• The speed button offers 0.75× to 3×. Speed changes apply at once and cost nothing, because the audio element does it.

Every action is also a command, so you can give it a hotkey: Play or pause, Listen from cursor, Stop, Next sentence, Previous sentence, Speed up, Slow down, Choose voice.

What is read

Headings, paragraphs, list items, quotes, callouts and footnote text. Links are read as their text, wikilinks as what Obsidian shows for them, and tags as their name.

Not read: properties, code blocks, tables, math, comments, images and embeds, link targets, bare URLs, footnote markers and block ids.

Limits

• Highlighting works in Live Preview, Source mode and Reading view. Reading view always starts from the top of the note, and footnote text is read there without a highlight.
• Live Preview draws callouts as a block of its own, so the highlight does not show inside one unless your cursor is in it. Source mode shows it.
• The first audio arrives about one to three seconds after you press play. After that the plugin renders ahead of the playhead and playback is continuous.
• Desktop only for now. Nothing in the plugin needs desktop, but it has not been tested on a phone or tablet, so it does not install there yet.

Network use and cost

The plugin sends the text of the passages you listen to, and nothing else from your vault, to api.speechify.ai with your API key. Speechify bills your workspace per character. A passage is sent when the playhead is about to reach it, a few hundred characters ahead, so a note you stop after one sentence costs about that much.

Rendered audio is cached on your device, outside the vault, keyed by the passage text, voice and model. Hearing a passage again is free, and editing one paragraph leaves the cached audio of the others in place. The cache holds up to 400 MB and drops what you heard longest ago.

Development

npm run dev        # rebuild main.js on change
npm test           # unit tests
npm run typecheck
npm run build      # production main.js

CLAUDE.md holds the rules the code depends on. RUNBOOK.md holds what has been verified and how to test the plugin in a throwaway Obsidian instance.

License

MIT