Hej, I am still working on FisherLoop [1] to learn Swedish (but I have added German, Spanish, French and Italian since). I created FisherLoop because I like audiobooks for language learning but I hated having to pause to look up words + I want to read along the book while listening. With FisherLoop I made "interactive audiobooks" where I use TTS with word level timestamps to highlight the words as they are spoken + I can click on words for the translation.
I am using cerebras for book translations and verb extraction and all LLM related tasks. For TTS I am using cartesia. I have played around with Elevenlabs and they have slightly natural sounding TTS but their pricing is too steep for this project. Books would cost a couple of hundred euros to process.
Ah, a Fortran joke. That's not just "technologists". That's old farts.
(In Fortran 66, variables didn't have to be declared. They would be integer if they began with I, J, K, L, M, or N. Otherwise it would be floating point [REAL, in Fortran parlance]. To this day it's why for loops usually use "i". With the bonus joke that God is real unless declared integer.)
Maths is the reason for loops use i. Fortran defines variables starting with those letters to be integers because maths has used those letters for iteration, counting, indexing, etc. for centuries. It was natural for a formula translating system to follow suit.
Probably the most appropriate one. The ESA is not quite part of the EU and has non-EU member states (including _Canada_), so .eu would be inappropriate (the EU is itself a member of ESA, but most EU member states are also members in their own right.)
Hej, I made FisherLoop[1] to learn Swedish. FisherLoop are interactive
audiobooks where I use TTS with word level timestamps to highlight the
words as they are spoken. This helps me pick up on pronounciation and
grammar in a, for me, natural way. Additionally, I added flashcards from
the books + word lookup. I am adding new books right now. If you have any
requests: public domain books, which are around one hour reading time let
me know :)
I am using cerebras for book translations and verb extraction and all LLM
related tasks. For TTS I am using cartesia. I have played around with
Elevenlabs and they have slightly natural sounding TTS but their pricing is
too steep for this project. Books would cost a couple of hundred euros to
process.
Did you use the hi@... email? I am seeing a hard bounce for that email. Not sure how to debug that right now. All my emails I have tested have worked. Could you try a different email while I debug?
I am using it for FisherLoop [1] to translate text/extract vocabulary/generate example sentences in different languages. I found it pretty reliable for longer paragraphs. For one sentence translations it lacks context and I have to manually edit sometimes. I tried adding more context like the paragraph before and after, but then I found it wouldn't follow the instructions and only translate the paragraph I wanted but also the context, which I found no good way to prevent. So now I manually verify, but it saves me still ~98% of the work.
If anybody wants to play chess in their minds, Lichess has a option to turn the pieces invisible [1]. It's not quite as hard as playing by voice as you have the move list you can refer to but it's a start.
There is also an option to have the computer read moves to you and you can make moves and query the state of the board by typing commands. Streamers use this to play blindfolded, while the stream shows the board position normally to viewers.
I created this as a tool for me to memorize chess openings. Since then I added more tools like tactics and endgames where I took theoretical winning endgame positions with the challenge to beat stockfish in them.
The project is open source [1] written in Elixir using Phoenix and also utilizing LiveView for some pages like the search pages.
Hah, I am doing the same thing on my chess site [1]. On my site you play against an opening repository so there is not even any real calculation done just a lookup into the repository what the next move should be. It feels really weird having the computer respond when you feel like you haven't even finished your move. That is why I added a 300ms delay [2] to make it feel more natural.
I am using cerebras for book translations and verb extraction and all LLM related tasks. For TTS I am using cartesia. I have played around with Elevenlabs and they have slightly natural sounding TTS but their pricing is too steep for this project. Books would cost a couple of hundred euros to process.
[1] https://www.fisherloop.com/en/