Hacker Newsnew | past | comments | ask | show | jobs | submit | sankde's commentslogin

This is an exciting milestone I have been working towards. I want to make Jexxa the most accurate transcription engine possible. Its automatic evaluation suite can lower the word error rate to 1 in 100. It automatically adapts to you. This is something that no other software on the planet can do. So I hope you guys love it.

https://www.jexxa.org


The job of AES (Automatic Evaluation Suite) is to go over your recordings and your outputs automatically and try to adjust the system so you get more accurate output in the future. Everything here is automatic, everything here is designed to be on device and private.


This is so weird, lol. I got the exact opposite advice keep building, and I kept building.


The problem is that you think when you kept building is great while people's reaction is that it's not that interesting. Nobody will buy your stuff just because you kept building. How many things you consumed this week that were made by people that just kept building?


Yeah a ton of great builds consumed. This is HN. There’s amazing bespoke tech everywhere and it’s oversupplied, that’s the real reason OP doesn’t get traction.

Just because you and the system of VCs and angels like a bro with connections and charm doesn’t mean we here do.


I'm a builder myself not a beggar. If you think you are entitled to an audience just because "this is HN", you are the one trying to be part of the "system of VCs"


Yeah, I corrected all the mistakes there. Are you sure, you can't open it? I have more than 10k visitors


You did not fix it, the landing page is still horrendous.


I did. The feedback was basically why are you doing this? I said I think so I can scale these models, efficiency and or make a lab of my own


Don’t stop pitching until you’re funded. Every time you pitch should be better than your last. Wishing you funding.


Yes I will keep pitching Thanks for your wishes


My email is [email protected]


Thanks. I don't want my data, especially voice and text to be sent to servers. That's why I built this.

I love my competitor's product, but the amount of private data I sent over was really uncomfortable to me. Especially when sometimes it included Phone Numbers and some private info.

At the same time, tools like mine will fall behind if I don't keep updating the models.


If you care so much about user privacy, why do you exfiltrate events named "dictation_started", "dictation_completed", "dictation_rejected", "dictation_blocked", "insertion_failed", etc to PostHog, with properties named: "audio_seconds", "latency_ms", "duration_ms", "install_id", etc?


Because anyone building a system needs performance metrics to understand how it’s behaving in aggregate / if they update the model? With the exception of install_id these all seem innocuous. Install_id is maybe needed anyway for purchase management. Probably could use a session uuid to track the other metrics at the cost of blinding you to being able to find individual customers with a bad experience which is probably much more important for a small player like this.


Apple provides APIs for doing this in a privacy-sensitive way, instead of exfiltrating data like this, carte blanche using PostHog.

https://developer.apple.com/documentation/metrickit


Thanks for your Feedback. More Privacy Improvements, Package Corrections and Attribution Incoming.


Yes These metrics are there to check system issues. Like Crashes and etc. Also I need to check Model performance too. However your data is yours, and this data will be anonymized too. I really want Jexxa to be private and secure, as I am a heavy user of it too.


Yes, I am training better models. On Device Dictation requires more resources than API based systems.

My competitor charges more and at the same time the data is being sent to someone else's servers.


Better models than… whisper, cohere, parakeet v2/v3? Did I miss the comparison on the webpage?

Is it just the models I’m paying for, or an app too?

I’ve been using anomalyco/HEX and MacParakeet… which are free and work great for my uses…


Yes, I have used Handy, a lot. Open Models need the system around them to keep getting better. Also faster.

The models you have quoted take a lot of RAM, especially on Mac. That's why I am optimizing my models too.


We also have better models in training. Also the underlying system would get faster, the models smaller over time along with grammar correction, in voice commands and better live preview models


The way this used to be handled is to sell new versions.


Yes. Newer models is the direction I am going in. New Models are expensive.


I was thinking of such a payment system. I would love to offer you that as I am still testing the pricing.

Edit: You also would get updates for life.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: