Anki HTML Exporter

Exports Anki cards to a browsable HTML page in which every card is rendered by Anki's own engine — a card in the export looks the way it looks in the reviewer. The same dialog serves that page live from the running collection, reads the cards aloud (Narrator, with an OpenAI API key), and plays them as a rapid presentation (Hypnagog).

Install

From AnkiWeb: Tools → Add-ons → Get Add-ons…, code 265861717. Or build the .ankiaddon with python3 build_addon.py and drop it on Anki.

Four ways to use it

Tools → Export to HTML…, or select cards in the browser and use the context menu. Pick a deck, tags, and optionally an Anki search (is:due, -tag:leech, added:30); the Content tab decides what a card shows.

The page

• Card sides — Auto (Anki's answer side, plus the question when the answer does not repeat it), Q + A, A or Q; for all cards, or per card.
• Reveal on click for cloze deletions and image occlusion masks; study mode shows the front only and reveals on space.
• Filter by deck, tag, note type, card state and flag, with counts; a search box over text, fields, tags and ids.
• Info / Details / Fields — header row, scheduling strip and raw note fields, each switchable.
• Shuffle, night mode, print at three densities; on a phone, swipe a card away.

Cards are mounted as they come near the viewport and released once past, so an export of a shared deck stays usable.

Output. A folder (index.html with media/, css/, js/; MathJax and jQuery only when a card needs them) or a single .html with media inlined. Media is copied straight out of collection.media — audio, video, fonts, SVG — and [sound:…] becomes a player. Remote http(s) media is left alone unless Download media referenced by URLs is on.

Browsing live

Browse live serves the collection to your browser while Anki runs: cards render as you scroll, pictures come out of collection.media, and the scope stays changeable — Filter… offers every deck, tag and note type, the search box is Anki's own, and the decks and tags the dialog was pointed at arrive as ticks there, so changing them replaces the dialog's choice. Optionally the page can be reached from your network; QR shows the address as a code that carries the scope you are reading. Anyone with that address can read the whole collection, so hand it out with care. The server runs until Stop or until the profile closes.

Narrator

Narrate opens the cards of the dialog as a narrated slide show. Each card is shown as the export shows it, a language model writes a short spoken telling of it, and text-to-speech reads it. This uses OpenAI's API: enter an OpenAI API key under ⚙ on the page. Each narration is a model call and a speech call — roughly one to two cents per card with the default models — and nothing is spent without a key. The key is kept in the add-on's config.

• Cards — the dialog's scope, then Filter… (decks, tags, note types, card states, flags) and any Anki search in the bar, exactly as in the live view; browser, learning, creation or random order.
• Playing — the front first, the back a few seconds into the narration or on Enter; the cloze asked on the card is marked. Length per card 10 s to 2 min as a ceiling: a thin card stays short. Speed 1×–2× with pitch kept.
• Language — the card's own, or one set under ⚙; the voice is directed in it. Thirteen voices to choose from.
• Image occlusion — the region under the mask is shown to the model, which reads the hidden label; the narration is about that structure.
• Ask — a microphone and a text line under the narration; the model answers from the card, spoken in the same voice.
• Phone — ▯ shows a QR code that opens the page over HTTPS on a phone in the same network.
• Export — ⇩ writes the scope as an audiobook (MP3, a chapter per card) or a video (MP4). Needs ffmpeg. The dialog says what it would cost first.
• Cost — a chip shows today's spend; the settings hold the counts, a daily budget, and the price list. Scripts and audio are cached in user_files/cache/, so a card narrated once costs nothing again.

Hypnagog

Hypnagog opens the cards of the dialog as a rapid full-screen presentation in the browser: each fact flashes for an instant, shows — a cloze target blanked and then revealed in its own colour, an answer fading in under its question — and fades out; after a round the deck reshuffles. It starts at once on the dialog's cards with the settings kept; Options (top right, or the O key) opens a sheet over it for the cards (due, new, all, leeches, failed in the last day or week), how many, how long each shows, the look (with or without CRT effects and ambient tones) and progressive speed. Needs no key.

Known limitations

• Image occlusion text shapes are placed without measuring the glyphs, so a label's plate can sit a pixel or two off; rectangles, ellipses and polygons are exact.
• Media that a template's JavaScript builds at runtime cannot be discovered and is not copied.
• Other add-ons affect the export only with Let add-ons post-process cards on, and hooks written for the reviewer may inject controls that do nothing outside Anki.

Requirements