Anki 23.10 Beta 1-4

Hi all,

A new beta is available. Please let us know how it goes for you.


First to test :slight_smile:

When using Image Occlusion to occlude an image, the color of the occluded part cannot be modified, which is not easy to recognize when adding occlusion to an image with similar background colors.

I didn’t find where to train the model directly inside Anki as the 23.10 Beta 1 changelog said.

After enabling FSRS, you can compute the optimal weights in the advanced section.

This button has no effect, clicking it just flashes a rectangular box in the upper right corner, too fast to see what it is!

You need to select multiple occlusions first.

Yes, I did. I recorded the action.

Use the button before it to group multiple objects into one group, and that button to ungroup the objects. Use shift to select multiple objects.

1 Like

It seems, that “Compute optimal weights” for FSRS is using my whole collection instead of only cards using selected preset.
How I tested:

  1. Created a new test deck.
  2. Moved only 1 card to the test deck - it has 5 reviews.
  3. Set test deck to use new preset, not used by any other deck.
  4. Clicked “Compute optimal weights”, it started computing weights for 150’000+ reviews.

When I added in the search criteria "deck:test" it completed very quickly with the expected message: “Only 5 reviews were found. The custom weights are likely to be inaccurate, and using the defaults instead is recommended.”


Thanks, it’s a bug, and will be fixed in the next beta. Use preset:"Preset name" until then.


When I run calculate optimal retention I get insufficient review history. While When I run compute optimal weights it tells me that I have 60k reviews.

Also do the settings applied also work for subdecks.

Still todo: Retention computation fails for v2 users · Issue #2667 · ankitects/anki · GitHub

1 Like

do the settings apply to subdecks or only decks that have that deck option

It seems it’s not possible to save free form cards in IO:


1 Like

I think these rows shouldn’t be included in this search result.

1 Like

Have you thought about how this functionality will be for updating the add-ons with the new nomenclature?

1 Like

Some other examples of drastic interval drop after activating built-in FSRS (85% retention, default weights)

The only pattern I noticed until know is that the cards affected by the interval drop all seem to 1) Have been “first reviewed” by manual reschedule and 2) Have 0 lapses

Examples of cards that are not affected by the interval drop (on the contrary, at least some of them seem to have very long intervals):

On closer inspection, I noticed that cards whose first “review” was a reschedule and that have never been lapsed (the ones that are affected by the interval drop), do not have Stability, Difficulty nor Retrievability

Some cards of the cards that were “introduced” using reschedule and have 1+ more lapses, also do not have Stability, Difficulty, nor Retrievability, but built-in FSRS gives very long intervals for them, e.g. this one going from 5.3 months straight to 14.6 years (45.8 years if choosing Easy)

It seems that some cards there were introduced “normally” also have “no Stability, Difficulty, nor Retrievability” and very longs intervals, although much more rarely


Some quality-of-life suggestions to improve user-experience:

  1. When I add a custom query for cards to train on, will cards outside the preset also be included? If so, I don’t want that, but tbh I forgot how to query for a preset.
    It would be nice if you include the query for the current preset into the grey hint text in the Compute optimal weights text field or add it even as a default, to make it easy to query for subsets.

  2. In Compute optimal retention, I was surprised it automatically set the newly computed retension without asking me. I would prefer if it would display the optimal retention in an seperate ideal retention fiel with an set/accept button to use it.

  3. When computing optimal retention, for Max seconds of study per day, it would be nice to be able do use the same time shortcuts as in the learning steps field for instance, e.g. 30m or 2h. Also the base unit could be minutes imo, not sure why that needs seconds resolution.

Btw, I’m using the git build for Anki c788f8ac, which contains some of the bugfixes alreadly, but I hope it’s not related to my UX suggestions.


No idea how I did this, but the optimizer is stuck
Speaking of which, why are there two Cancel buttons?