Help centre.

Content is Memojin's basic unit: a chapter, a lesson, a topic. Everything related to it lives in one place, across six tabs:

• Sources: where the content came from (document, video, pasted text, Anki deck).
• Lesson: the text, editable.
• Summary: the short version, generated from the lesson.
• Memo cards: the notions to remember.
• Questions: the exercises.
• Podcast: the audio episodes made from the lesson.

Content may carry only some of that. Empty content with just a title is perfectly valid: you fill it in later.

They have merged into My content. It was three filing systems for one thing, each with different rules.

• Your courses and chapters have become folders.
• Your lessons and decks have become content.
• Nothing was lost: cards, questions, summaries and sources followed, and your review history with them.

A deck had no lesson and no summary; a course refused standalone questions: those limits are gone. Every piece of content can now carry everything, or almost nothing.

Filing. A folder holds content and other folders, up to 5 levels deep.

A top-level folder carries an appearance — a colour, an icon or an image — so you recognise it at a glance. Subfolders don't: they inherit the parent's look.

A folder has no effect on memory: it neither speeds up nor slows down your reviews. It exists to help you find things, and to share in bulk — it is the folder you link to a space so that everything you drop in tomorrow enters automatically.

In My content, the New button offers three routes:

• Manual content: written by hand, with no source or AI. Handy when you take notes as you go.
• From a source: text, file, YouTube video, Anki deck or AI generation.
• Folder: to file what comes next.

In the first two cases you choose where to put it: at the root, or in an existing folder — which you can create without leaving the dialog.

PDF (the best-handled format), Word (.docx), PowerPoint (.pptx), plain text (.txt), and JPEG, PNG, WEBP or HEIC images — a photo of handwritten notes works, including one taken on an iPhone.

A few rules:

• Documents and photos can travel in the same import: photos are grouped together, the document keeps its pages.
• With several files (or a long PDF) you choose One content per file, A single content — the files are then read together, in the order shown — or Split into sections: Memojin proposes sections that you adjust before the content is created.
• A photo taken sideways can be straightened before it is sent: the rotate button on its thumbnail turns it a quarter turn per tap. It matters, because text recognition does not read a page lying on its side.
• Reading pages consumes your AI quota, in proportion to the number of pages.

Uploaded files stay available under My files, from the user menu.

No. If the tab closed or the connection dropped between the upload and the launch, your files did arrive but their processing never started.

A quarter of an hour later, My content shows a banner naming them, with a “Resume” button: the file starts over, one piece of content per file. Nothing was charged in the meantime — only page reading is, and it never happened.

You'll also find them under My imported documents (user menu), marked “Never started”.

Above a certain size, Memojin warns you and offers to process only part of the document: the first pages, the last ones, or a selection you make yourself, page by page, with a preview.

It is as much about time as about quota: processing a whole document can take several minutes and consumes pages from your AI quota.

For a full set of lecture notes, the best result comes from splitting it into separate pieces of content — one per section. That is exactly what “Let Memojin propose a split” does: Memojin reads the whole document, proposes sections from its headings, and you adjust them before creating one piece of content per section.

When you upload one or more files, choose “Split into sections”. Only page reading starts: it is billed like an ordinary import, the split itself costs nothing more.

When reading is done, the task queue shows “Split ready” and opens the split screen. Memojin proposes sections based on the document's headings, with an excerpt of every text block and every image.

On that screen you can:

• cut a section between two blocks, or merge two neighbouring sections;
• move a block from one section to another by dragging it;
• set aside a block or a whole section (an appendix, an administrative page): it will produce no content, but you can take it back;
• rename the sections and the folder that will hold them;
• choose, section by section or for all at once, what gets generated: summary, flashcards, questions. The cost shows as a percentage of your monthly quota.

“Create N items” creates the folder and one item per section, then starts the generations. Each item only carries its share of the document — cards are sharper, and the content stays readable. Until you click, nothing is created.

If the proposed split doesn't suit you, say so: the “What should change” field takes a free instruction (“one section per lesson”, “answer keys separately”), and “Suggest another split” runs it again with that instruction. You get three proposals per import; each replaces the previous one, so hand-made corrections are lost when you relaunch — Memojin asks first. It costs nothing more: only the pages read at the start were billed. And if the model could make nothing of the instruction, the screen says so: the proposal then comes from the document's headings again, and the attempt is used up all the same.

New → From a source → Text. Give it a title, paste your text (the « Paste » button reads the clipboard), and Memojin turns it into content.

This is the quickest route when your notes already exist elsewhere. By default the text becomes the lesson, from which you then generate the summary, cards and questions. You can also uncheck Write the lesson to get cards only — see « Can I create cards only, with no lesson? ».

Yes. New → From a source → YouTube, then paste the link. Memojin retrieves the video transcript and turns it into a structured lesson — unless you uncheck Write the lesson, in which case the transcript only serves to build the cards.

The original video stays available in the content's Sources tab, with a link to reopen it on YouTube.

If the video has no available transcript the import fails: that is a YouTube limitation, not a Memojin one.

Yes. New → From a source → Anki, then drop your .apkg file. Memojin reads the deck, tells you how many cards it holds, and imports them into a new piece of content.

The original .apkg file stays downloadable from the Sources tab.

Your Anki cards start a new life in Memojin: they enter the same memory as the others and are rescheduled by FSRS. Their Anki history does not follow — they start as new cards.

Yes. At the Generation step, uncheck Write the lesson: Memojin then writes no text at all and builds the cards straight from what you dropped in.

You get content with no Lesson tab, but which keeps its source (the PDF, the pasted text, the video) and its cards. That is what the old card decks were.

Two consequences:

• the summary is no longer offered: it summarises the lesson, and there is none;
• the import costs less, since no text formatting takes place.

At least one item must be ticked — the lesson, memo cards or questions: otherwise the content would be created empty.

New → From a source → AI. Describe the topic (« the French Revolution, causes and consequences »), add details if useful, and say how many pieces of content you want.

Memojin creates a folder and generates the requested content in it, with its lesson. You then move on to the summary, cards and questions.

Generation draws on your academic profile: explanations match your level. Everything stays editable — generated lessons are lessons like any other.

• Overview: the sheet at a glance — what exists or is missing, which spaces it lives in, and where to start. It is the opening tab.
• Sources: where all this came from.
• Lesson: the reference text. It feeds everything else.
• Summary: the key points, in short form. Useful to revise quickly before a test, and to generate tighter cards.
• Memo cards: the notions to remember, the ones memory schedules.
• Questions: the exercises, to check you can actually retrieve what you know.
• Podcast: the audio episodes made from the lesson.

An empty tab explains what it expects and offers to fill it.

It is the tab a piece of content opens on: the sheet at a glance.

• Learning material: one tile per part (sources, lesson, summary, memo cards, questions, podcast, notions). Each tile carries its verdict in large type — a number when there is one, a check mark otherwise (the lesson is written, the summary is ready) — and its tint lets you find it again without reading it. A dashed tile says what is missing and the gesture to create it; clicking a tile opens the matching tab. When a job is running on the content, the tile concerned shows it (a ring spins in place of its icon, a bar runs along its foot) and the writing gestures go quiet until it is done — that is what prevents two generations launched at once on the same material, or a lesson rewritten by hand while the AI rewrites it. Reading stays open: you browse the tabs as usual, and a banner above them says which job is running and that the gestures come back as soon as it is finished. A podcast or a soundscape blocks nothing: they do not touch the text, and only their own tile animates.
• How much of the lesson the cards cover: the share of the lesson that already has its memo cards, and how many to generate for the rest.
• In spaces: the list of spaces this content lives in, what is shared there (six small icons, lit or dimmed), and your progression and memory in that space — that is where you train, the “Train” button takes you there. On a class or an individual follow-up you steer, the button reads “Steer” and opens the space’s Steering — you read the class’s average progression there, never anyone else’s memory; if the sheet only lives in spaces you steer, the block is titled “where your students train”.
• Details and notions: folder, language — chosen at creation, fixable here, and it steers every generation (summary, cards, questions, podcast) —, attached voice, last edit, and the linked notions.

A still-empty sheet shows “Where to start”: a source or a lesson first, the rest is generated from it.

Open the Lesson tab and click Edit. The editor accepts rich text, headings, lists, tables, images and mathematical formulas. Three paths lead to the same blocks: the toolbar, the / key at the start of a line (type “/heading” or “/table”, for instance), and the + that appears in the margin when you hover a line — it acts on that very line.

Saving is automatic: above the editor, a marker reads “Saved” or “Saving…”. Click Done to go back to reading.

While reading, Show linked cards — shown as soon as a card was born from a passage — displays them in the margin.

On content shared by someone else the lesson is read-only: only its author can change it.

It is read-only too for the length of a job — reading the pages, writing, the summary, the cards, the questions, the notions: the Edit button turns off, and an editor already open goes read-only without closing, so that you do not rewrite the same text while the AI is rewriting it. A banner above the tabs names the job that is running; everything comes back as soon as it is done.

The summary boils the lesson down to a few points. It serves two purposes: revising quickly, and acting as a tighter base for card generation.

It is generated from the Summary tab, provided the lesson exists. You can edit it by hand, print it to PDF, and regenerate it — bearing in mind that regenerating overwrites your edits.

If the lesson changes afterwards, the summary is flagged as out of date: it no longer quite reflects the text. It is up to you to decide whether that warrants regenerating.

The summary does not count towards your progress: cards and questions do.

A memo card holds one notion to remember, written as a statement. It is not a question: it is the unit of knowledge memory works on.

This is what FSRS schedules: each memo card has its own memory state, its return date and its level of mastery.

A card can carry an image or a sound, and be linked to the passage of the lesson it came from.

A question checks that you can retrieve, not merely recognise. Eleven formats, from the easiest to recognise to the most demanding to retrieve.

Recognise

• Single choice — one right answer among several: you can tell close options apart.
• Multiple choice — several right answers to tick: you know a notion covers more than one thing.
• True / false — a statement to judge: you spot the flaw in a plausible sentence.
• Fill in the blanks — a text with missing words. Three ways to fill them: by typing, with labels (proposed words, decoys included, that you drop into the gaps) or from a menu (one list per gap). You put a term back into its sentence.
• Matching — two columns to link: you hold the pairs (word / definition, date / event).

Retrieve

• Open answer — you write, the AI marks and explains: you can phrase it, not just tick it.
• Reordering — items to put back in the right order: you hold a chronology, a procedure, a line of reasoning.
• Sort into categories — items to drop into bins: you tell close families apart.
• Short answer — a word, a name, a date to type: you retrieve the exact term unaided. Marking ignores case, accents and punctuation according to your settings; list the accepted variants, with and without the article (“Élysée”, “l'Élysée”).
• Numeric answer — a value to enter: you retrieve the right magnitude. You set a tolerance (absolute or as a percentage), and the unit is shown next to the field.
• Complete the table — a grid with empty cells: you restore several linked values (a conjugation, a classification table, a balance sheet).

All eleven formats accept an image. A question can be attached to a memo card — one card may carry several, at different difficulties — or stand alone with no parent card.

Questions feed training. They count towards your progress, alongside memo cards.

Yes. When creating or editing a card you can attach:

• an image (PNG, JPEG) — a diagram, an artwork, a map;
• an audio file (MP3, M4A, WAV, WebM) — useful for languages and music.

The attachment is shown during the session, before you answer.

You don't always have to do it yourself: when a card comes from an imported lesson, Memojin automatically attaches the figure from the section it was drawn from — the map, diagram or chart the card comments on.

From the matching tab of the content, or right at import — the wizard asks what you want to get: the lesson, a summary, memo cards, questions. At least one must be ticked; whatever depends on something else appears indented, and goes dim when its condition falls.

You choose the basis:

• From the lesson: the full text, so more cards and more detail.
• From the summary: shorter, already filtered, so tighter cards.

The AI picks each question’s format among the eleven available, and favours those that make you retrieve (short answer, numeric answer, reordering, sorting into categories, completing a table) over those that only make you recognise (multiple choice, true / false).

Generation goes into the queue: you can close the tab and you'll be notified when it's done. It consumes your AI quota, and the estimate is shown before you start.

Yes, all of them — as long as no job is running on the content:

• Create a card or a question by hand, without the AI.
• Edit the text, the difficulty, the attachment, the answers.
• Delete a card: its attached questions go with it. A standalone question is deleted on its own.

While a material job is running on the content — reading the pages, writing the lesson, the summary, generating cards or questions — those gestures go quiet: the buttons turn off, and a banner above the tabs names the job that is running. They come back as soon as it is done. An edit drawer that was already open does still save, though: work in progress is never lost.

An image can be added to any type of question — all eleven — and it is kept when you change a question's type.

Deleting a card also erases its memory history: that is final.

A question is only saved if it is playable: a fill-in-the-blank needs at least one blank and one answer per blank, a matching question at least one pair, a multiple-choice question at least two options. The refusal names what is missing — better to read it while writing the question than to fail it during review.

Yes. From the Memo cards tab of a piece of content: Add cards → Import from content. Search for the source content, tick the cards you want, and they are copied into the current one.

They are copies: editing them here does not touch the original.

The same menu holds Import from Anki, which opens its own dialog: the two imports are no longer mixed together.

While a job is running on the content — reading the pages, writing, the summary, generating cards or questions — the Add cards button is off: both imports wait for the job to finish, and the banner above the tabs names it.

A menu appears offering four actions, applied to the selected passage only:

• Explain simply
• Give an example
• Summarise
• Create a card from the passage

The first three open Jin with the passage in context. The fourth directly creates a memo card, linked to that spot in the lesson.

Where the content came from, whatever that is:

• the imported document, viewable and downloadable;
• the YouTube video, with a link to the original;
• the pasted text;
• the Anki deck, downloadable as .apkg;
• the generation prompt, if the content was written by the AI, with its date;
• the hand-made cards, counted separately.

In a shared space the sources are listed but the original files are not downloadable: we say where the content came from, we don't redistribute it.

Yes, whichever it is. Nothing in the application knows the name of your subject: if your material fits in a PDF, a photo or a few paragraphs, it fits in cards.

What changes from one subject to the next is the shape of the cards. Mathematics and physics need formulas, typeset as such inside the card; biology and geography need image-backed questions with markers to name; languages need cards you can hear; history and literature need fill-in-the-blank that demands the exact date or word.

Twelve public pages show this subject by subject, each starting from a real excerpt.

No, and that is deliberate. Memojin offers no catalogue of lessons: you upload your own documents — your teacher's handout, your notes, a photo of the board — and the application turns them into cards.

So you revise exactly the syllabus you will be examined on, not a stranger's version of it.

From the Podcast tab of a piece of content. It is made in two stages:

1. Writing the script — about a minute. You can close the tab, it will wait for you. If demand is high, this step may first indicate it's waiting, other jobs going ahead of yours: nothing is lost, the episode keeps its place.
2. You proofread and correct the script, line by line: this is the only window where changing a sentence costs a sentence, rather than a quarter of an hour of remaking.
3. The voices record. A long episode can take up to fifteen minutes; here too, a queue can delay the start without losing anything. You are notified as soon as it can be listened to.

Several lengths are offered — capsule, brief, narration, episode. Each says whether it plays as a two-voice dialogue or a narration. A lesson that is too short leaves the longer lengths greyed out and locked; hovering one gives the reason.

A transparency notice, mandatory, states that the episode was written by AI. It cannot be removed.

While you listen, the script follows the voices: the line being spoken lights up in the Podcast tab, and clicking any line jumps the playback to that passage.

If production fails, the tab and the generation queue say so plainly. Failure while writing the script: pick a duration and relaunch. Failure while recording the voices: the script is kept, you only relaunch the audio production from the review screen — lines already recorded are not redone. On a technical failure, try again later or contact support if it persists.

By default, synthetic voices: a host and a specialist for two-voice formats, a single voice for narration.

A teacher can apply their own voice, reproduced by AI, to the episodes and cards of their content. The episode then says so explicitly: « Voice of …, reproduced by AI ».

Making an episode is in beta: it is not deducted from your AI quota.

Yes. When creating an episode, tick “Sound ambiences”: a subtle background bed — a forest, a market, a classroom — is laid under the narration wherever the topic calls for it. The bed automatically ducks whenever a voice speaks, so it never competes with the explanation. If nothing in the catalogue matches the subject, the episode is simply made without a bed — that is a deliberate choice, not a failure. The box is unticked by default, and an episode that already exists is not rebuilt.

The My content toolbar offers:

• a sort: alphabetical, recently modified, recently added;
• a view: list or grid;
• a filter showing only content that has a lesson, a summary, memo cards, questions, a podcast or sources.

Filter checkboxes add up: content must tick everything selected to appear. The filter looks in the current folder and in all its subfolders — from the root, it goes through your whole library — and folders step aside while a filter is on, leaving only content. It is the fastest way to find what has no cards yet.

Through the row's action menu — the three dots — or, for open content, the menu at the top right.

Move to… opens your folder tree; Rename changes the title. A move is refused if the destination would exceed the 5 levels of depth, or if you try to put a folder inside itself.

With the mouse too: drag a row onto a folder in the list to file it there. During the gesture, a Drop into strip appears above the list with the root and the parent folders: drop the row there to move it up. Dropped outside any target, it stays put.

Moving content changes nothing about its memory or its sharing: it stays visible in the spaces it was added to.

Deleted content leaves your content and every space it was shared in. Its cards and questions go with it, along with their memory history. This is final.

A deleted folder takes its subfolders and the content it holds with it.

The confirmation dialog tells you each time what is about to disappear. If you only want content to stop being visible in a space, don't delete it: remove it from the space, which is an entirely different act.

Yes. The Select button in the toolbar shows the checkboxes. A bar then appears at the bottom with the number of items selected and the available actions: Move or Delete.

If an operation fails partway, Memojin tells you how many items were processed before it stopped — nothing is left in an ambiguous state.

Two routes to the same result:

• from My content: action menu → Add to a space;
• from the space itself: Content tab → Add content.

You choose what you share, where to place it in the space, and which parts go in (the lesson, the summary, the cards, the questions, the podcasts, the sources).

The content stays with you: it is shown in the space, not moved.

From the user menu → My files. You see every uploaded document, its date, its processing state and the content it belongs to.

It is also where you check that an import finished properly, and retrieve a document you no longer have to hand.

Every generation — lesson, summary, cards, questions, podcast — goes into a queue reachable from the header. It shows the current step, the progress, and the result once finished.

You can close the tab, shut down the computer, come back later: the work continues server-side. A notification tells you when it's done.

If a generation fails, the queue says so and you can restart it. An interrupted generation does not consume the quota it did not use.

When a single action starts several contents — for instance “one content per file” —, the queue groups them under one expandable card: “3 contents being created”, then “3 contents created”, with a button to open the folder.

Why is every button greyed out? Because a material job is running on that content: reading the pages, writing the lesson, the summary, generating cards or questions. For as long as it lasts, every writing gesture goes quiet across the whole sheet — generating, writing, creating a card, editing it, deleting it, reordering, importing, changing the language. Two writes at once on the same material is a conflict: one of them would be lost without warning.

Three signs say so:

• The banner, above the tabs: it names the job that is running — “Writing the lesson”… — and reminds you that the gestures come back as soon as it is done.
• The ring, on the tile concerned in the Overview tab: it spins in place of its icon, and a bar runs along its foot when the progress is known.
• The tooltip, when you hover a button that is off, which says the same thing again.

Reading stays open: you browse the tabs, filter, search, listen to a voice-over, start a session. And an edit drawer already open when the lock falls does still save — work in progress is never lost.

A podcast or a soundscape blocks nothing. They do not touch the text and they are long — an episode takes up to a quarter of an hour: only their own tile animates, and the rest of the sheet keeps being written.

Yes, but on demand: automatic production has been removed. From the Memo cards or Questions tab of a piece of content, the Generate voices button produces audio for cards that don't have any yet — with the synthetic voice, or a teacher's cloned voice applied to the content. Each card is read in its own language — French, English, German, Spanish or Italian — by that language's synthetic voice.

During a session, a button mutes or unmutes playback at any time.

On a piece of content you own, in the Memo cards tab and in the Questions tab, the Generate voices button appears as long as at least one card has no audio.

This production is in beta: it is not deducted from your AI quota. It joins the generation queue, where it may first show "Waiting — other jobs go first", then progress card by card ("Audio: n of N"). The speaker icon appears on each card as it's done, with no page reload.

Each card is read in its own language (French, English, German, Spanish, Italian): the synthetic voice changes with the card's language.

Yes, if it already had one: editing the content, answer, explanation, or options of a voiced card remakes its audio automatically, as soon as the queue allows it — the redo waits behind any batch already running on the same account.

A card that had no audio yet stays silent after an edit — nothing is produced until you use the Generate voices button.

It shows your content as a mirror of your memory: its headings, the concepts filed under them, and for each one how well you remember it — known, to review, or not seen yet. There is nothing to do: it draws itself from the concepts of your memo cards and appears as soon as the content has some.

Green: you've got this. Red: to review. Grey: you haven't worked on this notion enough yet for us to tell — it isn't a zero.

No, and that's on purpose: the map isn't a separate document, it mirrors your concepts and the cards linked to them. To change it, work on the material: rename or delete a concept, or move a memo card to another concept, from the "Memo cards" tab. The map follows on its own the next time you open it — no button, no rebuild.