The future of SRS - Anki with AI?

Anki Suggestions
1

I’ve been using Anki now almost a quarter of a century on and off with it’s different possibilities, mainly for language learning.

Up from some point HyperTTS and LanguageTools made it much easier and smoother for me, to create audios for my own and customizatized cards, compared to the old AwesomeTTS.

I hope Anki as a system will stay with me in one form or another from a purely technical point of view. These days.

But unfortunately Anki always was lacking the active side of actual speaking.

Online teachers were too complicated, impractical and in the long run ultimately too expensive for me.

So I’ve been testing AI-based apps for a long time now, but I’ve never been satisfied. Even ChatGPT itself I had to prompt hard to make it usable at least as a conversation partner. But it still misses the customized memory and tools for in depth analysis to act as a “teacher”.

I recently discovered “Speak” and that was THE revolution for me. At last someone is already using AI so that a foreign language can be learned and practiced simply by speaking.

They also have an SRS approach, but of course not as individually usable as in Anki.

I don’t think they’re as far along as I hope that there will soon be providers for this:

An AI tutor and conversation partner who always evaluates the level of my knowledge and skills in the background and always adapts topics and vocabulary in conversations and SRS learning. They are on the right track with “Speak”. But at the moment I still need my beloved Anki in parallel.

At the moment, I’m getting more and more examples of speech patterns from the AI to feed Anki. And it is great for my understanding, to consult dictionaries but discuss this topics with AI (meanwhile most with the integrated tutor of “Speak”). But so far still I don’t know of any AI SRS app that systematically analyzes, trains and expands my vocabulary beyond beginnings (1000 words etc.).

What developments are there, or have you at least heard of, that are really exciting for language learning right now. Especially of course with regard to Anki and SRS.

As I said, “Speak” was a real revelation because it got me speaking and keeps me doing that.

But I would also like to have an AI Anki, especially for analyzing and systematically expanding my own vocabulary.

1 Like
2

Not a development, but we should be able to connect Anki to external readers and systematically update the memory state of cards every time the user encounters a word in input.

1 Like
3

me… layman developer? rely on cursor ide + LLMs.

currently there already a addon called “anki connect”, that the otherside could
throw things at anki and it will become an anki card.

for the otherside you do need a program or something.
most famous one i think is obsidian_to_anki addon.
actually there are some esp chrome browser extensions that let you highlight words and throw to anki and become a card.

anki and the addon have done enough,
either it’s the other side have to do something,

or, the USER have to do something.

if you dont code, you can do like me, using cursor ide + LLMs,
or completely free-ish VScode + cline/copilot + LLMs to do so.

for LLM, claude 3.7 sonnet + thinking gives the best result.
others… chatgpt, C3.5, deepseek, grok3 are all ok.

google’s gemini is rather rubbish all the time. i dont know about their 2.0 version,
i really dont have patient in them.

1 Like
4

Thanks for your reply. At the moment I use ANKI and SPEAK side by side when reviewing. I have enough cards mainly sorted for frequency for two or three lives, I’m afraid. So I don’t need new mass import, rather the evaluation, what I need and what not. I had pampered around with the FSRS, but for that my cards vary way to much. With the old alghorythmus I better get along. As said, my wish would be a permanently assessing AI instead. SPEAK has a for language pre-prompted tutor chat (not quite sure, what LLM) I ask all the things I need, when reviewing. So I don’t use any other LLM with subscription, except from time to time perplexity again. On disadvantage on SPEAK, I have again and again to reprompt the formats, I want the output from the tutor in. There, it, doesn’t remember it. But for the time being I get along.