Would love to hear if anyone has had success with incorporating any of the generative AIs, whether it be ChatGPT, StableDiffusion, DallE, Suno, Udio, or whatever else.
Personally I have had some mixed success, but I definitely find it fascinating and fun to experiment.
Doug
I use ChatGPT to create a sentence mnemonic for the acronym I give it then give the sentence to Flux to create an image of it.
I used Suno to create a song to memorize a text verbatim
I don’t know if it’s creative…
You can ask chatgpt (or I suppose any LLM) to generate lists in CSV format, to import into anki.
I did one feeding it a most frequent words list, asking it to add in a second column the IPA with stressed syllable.
I also gave him a short story and asked him to extract the words which are also in the most frequent words list, and to generate a CSV file with the following columns: English, IPA, Context sentence, Translation.
Another thing I tried was to ask it ideas to fill up my 999 PAO list, but either I didn’t know how to ask it the right way, either it isn’t able to really know how words are pronounced (instead of just written).
For example, for repeated sounds: it kept proposing me coffee or cave for k-f-f. Even after explaining that there is only one f sound in coffee, it kept doing the same error.
I use Generative AI now too… both ChatGPT and Grok2 (built into X and part of the premium membership package). I hit snags with asking for things like asking for specific celebrities or cartoon characters due to copyright - sometimes it’ll produce the images and sometimes it won’t.
I’ve also found that it tends to be ok with rendering just one or two images at a time, but is hit or miss rendering images with three, and almost always screws up when asking for four details.
My workarounds include apps that erase backgrounds (turning the backgrounds transparent), a photo combiner (combines panels of photos in the order I need them), an app that removes objects (fills in the removed object with surrounding simple patterns or colors etc). By doing that I can paste my celebrity and cartoon character heads on the AI images, and add other details that AI gets confused with.
I use another app to annotate one copy of the completed combined image, and keep one copy without anything in writing (for training purposes).
Overall I’m pleased with the results BUT if AI were my Camera Crew, Set Designers, and Actors, where I am the producer and movie director - AI is not quite doing a great job. The uploads are an example of the meatball surgery I have to go through - you haven’t seen the cutting room floor that’s knee deep with discards.
Even so, I’m very happy using Generative AI augmented by my own hacks and talents.
Ok chatgpt is definitely not good with sounds. I asked him for words beginning with the sound /tɪ/ and it gave me tiger and ticket.
As for images, I’m wondering if it’s worth it (because my imagination is just much faster). Also won’t having to imagine the scene myself every time help with remembering? Like having to produce something yourself helps more for remembering than just re-reading it.
Of course if you want to share your system with other people it would be good to let them see what you imagine (but then not good for them to remember because their brains haven’t done the work of imagining?)
ChatGPT is challenged by sounds to numbers transformations using Major System rules. I found that out the first day - but that’s ok. It’s easier to do that stuff in my head anyway
Each of us has different preferences, strengths, and weaknesses. For me the work is being done up front during the meticulous repeated renderings and corrections, and editing repeatedly “in post” after getting the best I can get out of AI and then doing the detail work myself.
I find that my own sketches in pen and colored pencil take longer to make and edit, and that they’re not as vibrant as AI renderings.
I also find that not drawing them out, and then having to re-imagine them after not having rehearsed for too long of a period (due to work schedule or other spaced repetition schedule interference) isn’t practical in cases where the image in my mind just refuses to “click”.
Plain and annotated renderings used as study sets keep everything fresh, portable, shareable and secure. However I still use pencil and pen, highlighter, even (waterptoof) crayon if that’s all I’ve got, to put drawings on my index cards while working.
I like this one 
