61–70 of 73 posts
Earlier quoted context omitted.
Why does it take a lot of data? Afaik you can select lower quality in youtube-dl but you don't even need video do you?
I know you can scrape only audio from YouTube with YouTubeDL but it’s somewhat annoying
youtube-dl -f bestaudio $URL
Hey HN! I've been working on this for a while now. While there are other on-premise solutions using older models such as DeepSpeech [0], I haven't found a deployable project supporting multiple languages using the recent RNN-T Architecture [1]. Please note that this does not achieve SotA performance. Also, I've only trained it on one GPU so there might be room for improvement. Edit: Don't expect good performance :D t…
Hi! What should you need to implement other language i.e. Italian or French? I mean: it's a problem due to the less of datas or what? Another question: could you use for example mozilla voice data to train/test?
Thats why there are dozen RNNT projects around, some of them more reasonable, some less, but most of them struggle to demonstrate even a good librispeech WER.
You have a long way to go.
Earlier quoted context omitted.
You're effectively experimenting on your child? Don't you think the repurcussions might be severe?
Every parent is, though many will deny it. The repercussions of doing so without intention and without admitting it are already severe. Every parent is likely to traumatize their child in some way, including the trauma of protecting them from trauma to the point that they don't learn to heal through it. I'm ok with intentionally experimenting and normalizing healing within our family. Also, it's already paying off tr…
I wonder if the drawback will be that he/she will be incredibly bored once released to the "normal" world and the slower development pace of contemporaries and may feel out of place and frustrated. It's the curse of the gifted.
Earlier quoted context omitted.
Hi! What should you need to implement other language i.e. Italian or French? I mean: it's a problem due to the less of datas or what? Another question: could you use for example mozilla voice data to train/test?
Data and compute are the largest hurdles. I only have one GPU and training one model takes 3+ days, so I am limited by that. Also, scraping from YouTube takes time and a lot of storage (multiple TBs). Mozilla Common Voice data is already used for training.
I was hoping to use it as a good training base, but the issues I encountered made me wary that the data quality would adversely affect any outcomes.
Earlier quoted context omitted.
Every parent is, though many will deny it. The repercussions of doing so without intention and without admitting it are already severe. Every parent is likely to traumatize their child in some way, including the trauma of protecting them from trauma to the point that they don't learn to heal through it. I'm ok with intentionally experimenting and normalizing healing within our family. Also, it's already paying off tr…
That's a good observation! I have no specific fears at all, since I was wondering what sort of "experimentation" it involved. I wonder if the drawback will be that he/she will be incredibly bored once released to the "normal" world and the slower development pace of contemporaries and may feel out of place and frustrated. It's the curse of the gifted.
We're working hard to keep from using judgmental/subjective words like good/bad, like/dislike, etc. We're also starting to incorporate Nonviolent Communication patterns and concepts. We use they/them pronouns instead of gendering them. If they want to do something, we strive to help them do it as long as they wont be maimed or killed. We ask them for consent before changing their diaper, touching them, picking them up, taking things from them, and performing medical/dental procedures on them. We've named them Uni Verse All. They wear whatever clothes they choose, no matter what gender they may seem created for. I'm genderfluid and do the same. I also shower once every 1-2 weeks, stopped using shampoo about a year ago and am about to stop using soap on my body, too. We aren't teaching them about property currently and may not ever, choosing to instead describe things as "living with" someone. I'm developing a spirituality with a component I call "radical ignorance," which is essentially a sort of Zen "beginner's mind." It recognizes that ignorance isn't an excuse, but a spiritual reason for doing things, which runs counter to the US's legal reasoning of "ignorance of the law is no excuse for breaking the law." My partner and I are intentionally staying out of the workforce, instead choosing to serve people in our community alongside Uni, which allows both of us to be available so they can have their choice between us. Anytime one of us is choosing to not let them do something without there being a safety issue, the other ideally defaults to helping Uni do what they want. When they get hurt, we bring their attention to the pain and teach them to mindfully experience it while breathing through it.
As for the drawbacks, we're currently designing a community anti-adultist homeschool model that focuses on collaboratively learning our needs over what schools typically teach and allowing the students (of age 0-200+) to choose the contexts for learning. So they'll probably have an interesting intergenerational peer group to blow past the world with.
I LOVE that you provided a sample application targeting the ESP32-LyraT! While the ESP8266/ESP32 get plenty of love on HN (and elsewhere) I think the ESP ADF (audio development framework) and various boards dev boards (Lyra, Korvo, etc) are really under appreciated and essentially unknown. I enjoy a Raspberry Pi, Jetson nano, Arduino, whatever as much as the next person but the seemingly endless stream of projects an…
Thank you for your kind words! :)
Is there an open-source or paid SDK/API that I can use to create a group voice chat mobile app with "live" transcription? Or something that can plug-in to a system like this? I looked at Twilio but they seem to only offer a means to do it on their VOIP/SIP product.
> open-source or paid SDK/API that I can use to create a group voice chat mobile app with "live" transcription? Or something that can plug-in to a system like this? Yes, Google, Amazon, Microsoft all offer streaming solutions (wouldn't recommend Amazon's however, might recommend Microsoft over Google). wav2letter from FB is the only open-source framework worth looking at, deepspeech is not a seriously usable framewor…
Earlier quoted context omitted.
> open-source or paid SDK/API that I can use to create a group voice chat mobile app with "live" transcription? Or something that can plug-in to a system like this? Yes, Google, Amazon, Microsoft all offer streaming solutions (wouldn't recommend Amazon's however, might recommend Microsoft over Google). wav2letter from FB is the only open-source framework worth looking at, deepspeech is not a seriously usable framewor…
Curious why would you not recommend Amazon... is it cost or something else.
Earlier quoted context omitted.
Data and compute are the largest hurdles. I only have one GPU and training one model takes 3+ days, so I am limited by that. Also, scraping from YouTube takes time and a lot of storage (multiple TBs). Mozilla Common Voice data is already used for training.
Thanks for sharing this project. What do you think of the data with Mozilla Common Voice? The random sampling I looked at a while back seemed pretty poor -- background noise, stammering, delays in beginning the speaking, etc. I was hoping to use it as a good training base, but the issues I encountered made me wary that the data quality would adversely affect any outcomes.