PeerTube Arch Linux je váš domov ve světě Fediverse! Sdílejte videa s přáteli a objevujte nové, to vše na platformě vlastněné českou komunitou Arch Linuxu a nadšenci open source.
This is where I could really use your help, if you're interested! - Help with training costs: I've been fortunate to receive a grant from Google's TRC program, but there are some additional expenses. Any contribution would be incredibly helpful. - Donating spare GPU power or Colab credit: Even a little bit would be a huge boost!
@nick This is more efficient and lightweight. Compared to Piper, this model is more responsive and requires less system resources. Also this is a modern TTS implementation, I referred to papers published in 2023-2024.
GOtcha. I might throw you a few dollars to maek it go, especially if you can come up with training data that won't have the pronunciation problems that existed with PIper.
@nick Currently I'm exclusively working on the model architecture, which will resolve some of the issues. I'll train on freely available, high quality datasets, but creating a new dataset from scratch is beyond my current resources. I'll leave this for later, and I can help anyone who wants to take up this task.
@nick A good dataset will definitely help a lot. Not only me, but any future developer who will work in this field. Also, a high quality dataset can establish a bridge between our community and academia, where major TTS breakthroughs happen. We give you our dataset to evaluate your models, and you allow us to use your great TTS model architectures.
@Musharraf :verified: Please, when preparing the dataset what's the difference between train and val? If I have single speaker recording what do I put in those folders?
@pvagner In machine learning, A dataset is usually split into two splits. The 'train' split is the larger, and is used as input for model training. The 'val' split is relatively small, and it is used for evaluating how the model performs during training.
@pvagner Given you have a list of wav files and the corresponding transcription, first you need to decide on the size of each split. Depending on the size of your dataset, you can split it 95%-5% or 99%-1% for training and val respectively.
@tomecki Simply put: Sonata is an inference engine that can theoretically drive any TTS model. OptiSpeech is an actual model that generates speech, and the quality of the output depends on it.
I' afraid I cannot. The whole point of this and Sonata is to create high-quality, but very efficient and lightweight neural TTS. Coqui-TTS models are neither efficient nor lightweight.
@luiscarlosgonzalez Tortoise TTS is too heavy for a high-end server, let alone a standard computer or a mobile device. This system is designed specifically for running on a standard CPU.
@rperez030 Thanks for your contribution. Really appreciate it! For this month, Pneuma Solutions provided GPU resources for training. But I still accept donations to ensure we have enough resources for future training needs. I'm happy to receive your donations via my colleague's paypal address: paypal.me/geotts or Email: info@geotts.ge Please mention mush42/tts in the transaction note.
Please vote for an implementation of VJOURNAL in Thunderbird. 👇 This would REALLY improve the ecosystem and usability of open protocols! Currently there is almost no software on the desktop to sync your notes and journals with a CalDAV server and TB could fill a huge gap here.
# already has a great support for VTODO VCALENDAR and contacts. What is missing though is support for VJOURNAL entries. Journals are close to calendar entries.
You have to enter your password using this *very* Frenchy keypad. You don't have lowercase letters, the blanks are not spaces but just non-clickable gaps, but as a compensation you have some weird symbols that your keyboard does not have a key for (e.g. `µ`).
You have to enter your password using this *very* Frenchy keypad. You don't have lowercase letters, the blanks are not spaces but just non-clickable gaps, but as a compensation you have some weird symbols that your keyboard does not have a key for (e…
#DisabilityIN The Evinced CEO referred to the W3C ARIA Authoring Practices Working Group as one of the three pillars of creating accessible content. Using it can guarantee “100% accessibility”.
a single technology automatically making it possible to guarantee “100% accessibility” is at least as ridiculous as a single technology guaranteeing 100% security. This doesn’t just apply to automation/overlays; it applies to every technology used on any nontrivial project.
Replacing the word “accessibility” with “security” and vice versa is a good bullshit detector for marketing material.
Ultimately, nothing short of allowing the user to remake the entirety of the program's human-interaction layer to their liking will provide 100% accessibility.
#DisabilityIN Luncheon Plenary: “Nothing About Us Without Us: Accessibility Innovation, Leadership, and Impact”
Microsoft’s CAO discussed one challenge of #accessibility is non-practitioners creating accessible content. She talked about advances in tooling to solve that, using this slide.
#DisabilityIN Very quick demonstration of Microsoft Accessibility Insights run against the Disability:IN home page only. Found no issues! Applause from audience. I *think* the “Needs review” tab listed two issues.
#DisabilityIN The Evinced CEO referred to the W3C ARIA Authoring Practices Working Group as one of the three pillars of creating accessible content. Using it can guarantee “100% accessibility”.
Dear AppleVis Community, At his request, I wanted to provide you an update on AppleVis' Founder and Editor-in-Chief, David Goodwin; as well as the future of AppleVis.
@DavidGoldfield i saw him post on this topic yesterday. He was very thankful for all the support. He said he's on the road to recovery. I'll check and see if there's any further updates.
Recently, user Jordan shared a small but meaningful experience made possible with #NVDA: "When NVDA by NV Access reads alt text from your favourite #restaurant menu, now I just found a new #dessert I want to try! Thanks to NVDA, my day is getting off to a perfect start."
Hi !Friendica Support, I'm not finding much documentation about following and connecting. Is this how it works: If the only relationship between A and B is that A follows B, then: - A is a fan and follower of B. - A and B are not considered connected. - There is a unilateral friendship between A and B.
If B also follows A, then: - A and B are considered connected. - There is a mutual/bilateral friendship between A and B.
When you receive a friendship notification from another Friendica member, you will have the option of allowing them as a "Follower" or as a "Friend". If they are a follower, they can see what you have to say, including private communications that you send to them, but not vice versa. As a friend, you can both communicate with each other.
How do you cause a “friendship notification” to be sent? What form does it take when it arrives? I've been experimenting with 2 accounts, but have not seen a friendship notification.
Hi Dirk. It depends on the account type and network. For example my Fendica account is setup to accept all friends requests as "followers". So I get no notification when someone adds me as a friend. But if you follow a Mastodon account, and yo are on Friendica, perhaps they get a notification.
In Friendica you can tweak these things a lot. Let me know if you have more questions.
1. How do I set up my Friendica account to accept all friend requests as "followers"? And what are the other options?
2. I told my sister to take these steps in order to follow me: a. Go to Contacts b. Type my info in the “Add new contact” field and click Connect c. On the page that comes up, click Submit request She followed the instructions, and without any input from me, we are now mutual friends. Why didn't Friendica ask for my approval?
@Dirk Maas @Tio Go to settings>account>advanced account type. Unless you have standard, you wont need to approve everyone if they want to follow you. love-all will make them all friends, soapbox will make them all followers.
Thank you @anubis2814! I just checked and my account type is standard. But with my sister it acted like it was love-all. I'm not sure how to explain that, but I'll keep an eye on things going forward.
I see. So you didn't get a notification when she added you as a friend/follow but when I did you received. Interesting. Maybe was it that you already had her in your list of friends? Just checking...
No, she was not on any list – I knew she had an account, but that's all. I had tried to find her account, but could not.
Two unexpected things happened: (1) She became a follower of mine without any notification to me or input from me, and (2) I became a follower of her – again without any action taken on my part.
Yes @anubis2814 is right. You can change it form there.
It depends what kind of account you have. Maybe you have it setup to accept all friends requests by default. Please check that. And by the way to add someone as a contact is even easier: use the top search bar and paste someone's profile (url or name). Enter. Now you can submit a friend request. I just clicked to follow you did you get any notification?
@Dirk Maas As far as I've ever seen its just A follows you. Following and friendship are the same thing. Though you can both follow and ignore someone too.
Perry Johnson asks RNC attendees if they are better off today than they were 4 years ago and then immediately pivots to talking about 2019. I wonder why he skipped 2020 🤔
If, like me, you've put off upgrading from a gas range to an induction electric range, I have some fantastic news. In most of America, you can now pre-order a battery-equipped range, which just sips from your normal household outlet, and can work during a power outage. #ElectrifyEverything copperhome.com/products/charli…
The world's first energy storage equipped induction electric range. We’ve sold out of First Edition Charlie ranges, and are now accepting pre-orders. Your pre-order deposit secures your future order. It is fully refundable at any time.
Discover how to automate cloud backups for critical data using AWS tools like Lambda, S3, and CloudWatch, ensuring reliability and security. #DEVCommunity dev.to/emma_olatujoye/automati…
Gus Gus perform "" live at the KEX Hostel in Reykjavik during Iceland Airwaves '11. Recorded 10/12/11.Setlist:SelfossArabian HorseOverSound: Aron ArnarssonCa...
PyGObject, the #Python bindings for #GTK, have recently merged support for asynchronous programming, closing an orthogonality gap between Python's async support and GTK having been single-threaded / event based since forever: gitlab.gnome.org/GNOME/pygobje… This ends a decade of using projects such as gbulb that kindly bridged that gap.
I'm not sure he is on the fediverse, but huge kudos to long time GNOME contributor Benjamin Berg for this. This is such an amazing gift that he has worked on for the last couple of years and it really levels up PyGObject. 🤩 #Python #GNOME
Is Ohio Ballot Access Deadline a Reason for Democrats to Push through a Virtual Roll Call Vote Before the Convention to Choose Joe Biden? Almost Certainly Not electionlawblog.org/?p=144570
After a warning from the Ohio Secretary of State that the Democrats’ late convention would come too late to notify Ohio to put its nominee on the ballot, Democrats decided to hold a virtual roll call vote early before the … Continue reading Is Ohio B…
Okay, #Emacs users, does anyone still use Helm? Gonna get back into it, and setup is kinda different, so yeah. I loved it though when I had it working last time.
@Devin Prater :blind: Perhaps emacspeak has some code to handle helm auto completions. I think with speechd-el I did not managed to use it properly. Or that might be just me not being clever enough. I'll try looking into it again.
Tech firms are using troves of content they don't own to train their AI models. One dataset includes Taylor Swift lyrics, MrBeast videos, and even two clips from PCMag. pcmag.com/news/nvidia-apple-ai…
#GUADEC2024 starts this weekend! Are you ready? Make sure you're registered and be sure to check your email for event updates and links. Haven't signed up yet? There's still time, let us know if you're attending in Denver or online here: events.gnome.org/event/209/
It was this way across the company, Alejandra found, her phone lighting up with backchannel gossip as she walked. Big round of layoffs that also vaporized whole layers of middle management, and plenty of executives as well.
Have you been thinking about getting a new Apple watch? Right now, you can get an Apple Watch Series 9 on sale for nearly $120 off. This Apple watch is 28% off the normal list price of $429, and right now can be found for just $309.
Resubmitted: Complete Text of Article: 1Password product enhancements [Summer edition]: Recovery codes, auto-save, and more groups.io/g/tech-vi/message/73…
Element Web is completely broken with Firefox ESR. I don't blame the devs here, they didn't make the policy. But it baffles me that management apparently decided they simply don't support that and won't even do anything to make sure affected users can migrate their existing chats (as not everyone will have a key backup or a second session to allow verification).
This is e.g. the default browser on Debian stable and this is affecting a lot of users that had no idea they were using an "unsupported browser" until it just suddenly broke. It's one thing to not support Firefox ESR (though I think that would still be the wrong choice). It's another to just leave users that could not have reasonably known that without any kind of fix.
Your use case It seems (given the discussion in #27682, unless the person in charge of support there) that Firefox ESR isn't supported. Given it has a large user base (enterprise, Debian stable and...
This is a repository of apps to be used with F-Droid. Applications in this repository are official binaries built by the original application developers, taken from their resp. repositories (mostly Github, GitLab, Codeberg).
Před několika měsíci jsem začal přispívat do OpenStreetMap a celkem mě to chytlo. Rozhodl jsem se tedy napsat seriál článků o tom, v jakých formách lze přispívat. Seriál ale musím začít tím, že napíšu o tom, proč vlastně přispívat a kde všude může být vaše práce vidět.
Priznávam sa bez hanby, toto som vôbec netušil. A keďže Google aktívne bojkotujem, využívam Apple maps, v aute navigáciu od Kie, čo je ak sa nepletiem maskovaný TomTom a historicky zaplatený Sygic si držím ako zálohu... tak sa neviem dočkať ďalších článkov. Alebo ak máš po ruke nejaký slušný resource ako začať, tak sem s ním. 👍
Návodů bude myslím hodně, jde o to, v jaké formě chce člověk přispívat. Může to být StreetComplete, což je fakt jak sbírání pokémonů, jen jdeš a pár klepnutími doplňuješ chybějící info jako povrch cesty apod. Další level je doplňování POI z mobilních aplikací jako OsmAnd. Chybí hospoda? Doplním. A pak je mapování v plnohodnotném editoru, kde zakresluješ chybějící cesty, chodníky atd. To už vyžaduje určitě znalosti. V seriálu nemám ambici napsat ultimátní návody, ale spíš napsat o tom, co která forma přispívání obnáší, aby si lidi udělali představu, co by je mohlo bavit, jak moc se do toho ponořit.
Medzičasom som skúsil OsmoTagger a stiahol Go map!! Takže sa skúšam hrať. Myslím, že sa natrénujem na Vesničke mé strediskové 😂 Teším sa na články, vážne.
na seriál se těším. Mám chuť přispívat, ale nevím z jakého konce to vzít, tak snad mě to pomůže nasměrovat. Zatím jsem ve stádiu chytání pockemonů ve #SteetComplete a editace pár detailů v okolí mého bydliště.
Záleží co tě zajímá případně co tě trápí. Pokud si na něco netroufáš, nebo nevíš jak danou věc opravit, můžeš se zeptat. Buď tady (pár nás tu je - já třeba sleduji tagy #OSM a #OpenStreetMap ) případně máme emailovou konferenci a propojené místnosti jabber/matrix/irc/telegram/discord
Intel can't stay silent for much longer ("...Intel needs to address the instability on 13th-gen and 14th-gen CPUs. We’ve seen BIOS updates, comments from PR managers, and documentation shared with motherboard vendors...") digitaltrends.com/computing/in…
Google Pixel 9 could steal this upgrade from iPhone 15 Pro ("...Thread radios offer support for the Thread network stack— a mesh networking standard built for Matter-compatible smart home products. Thread is supposed to use significantly less power than Wi-Fi or Bluetooth, which should help minimize whatever impact controlling your smart home might have on your phone’s battery life. ") tomsguide.com/phones/google-pi…
Münchner Aktivisten hebeln die bayerische Bezahlkarte für Flüchtlinge aus. Die CSU spricht von "Asyl-Gutschein-Betrug". Wie der Trick der Aktivisten funktioniert.
Musharraf
in reply to Musharraf • • •- Help with training costs: I've been fortunate to receive a grant from Google's TRC program, but there are some additional expenses. Any contribution would be incredibly helpful.
- Donating spare GPU power or Colab credit: Even a little bit would be a huge boost!
Musharraf
in reply to Musharraf • • •All of my work on neural TTS is completely free and open-source.
github.com/mush42/sonata-nvda
github.com/mush42/sonata
Together, we can make high quality TTS technology a reality for more people.
GitHub - mush42/sonata-nvda: This add-on implements a speech synthesizer driver for NVDA using neural TTS models. It supports Piper
GitHubMusharraf
in reply to Musharraf • • •This means it is efficient and has low latency.
It is open-source:
github.com/mush42/optispeech
GitHub - mush42/optispeech: A lightweight end-to-end text-to-speech model
GitHubPratik Patel
in reply to Musharraf • • •Nick Giannak III
in reply to Musharraf • • •Musharraf
in reply to Nick Giannak III • • •This is more efficient and lightweight. Compared to Piper, this model is more responsive and requires less system resources.
Also this is a modern TTS implementation, I referred to papers published in 2023-2024.
Nick Giannak III
in reply to Musharraf • • •Musharraf
in reply to Nick Giannak III • • •Currently I'm exclusively working on the model architecture, which will resolve some of the issues.
I'll train on freely available, high quality datasets, but creating a new dataset from scratch is beyond my current resources.
I'll leave this for later, and I can help anyone who wants to take up this task.
Nick Giannak III
in reply to Musharraf • • •Musharraf
in reply to Nick Giannak III • • •A good dataset will definitely help a lot. Not only me, but any future developer who will work in this field.
Also, a high quality dataset can establish a bridge between our community and academia, where major TTS breakthroughs happen. We give you our dataset to evaluate your models, and you allow us to use your great TTS model architectures.
Peter Vágner
in reply to Musharraf • •Musharraf
in reply to Peter Vágner • • •In machine learning, A dataset is usually split into two splits.
The 'train' split is the larger, and is used as input for model training.
The 'val' split is relatively small, and it is used for evaluating how the model performs during training.
Musharraf
in reply to Peter Vágner • • •Given you have a list of wav files and the corresponding transcription, first you need to decide on the size of each split.
Depending on the size of your dataset, you can split it 95%-5% or 99%-1% for training and val respectively.
Musharraf
in reply to Peter Vágner • • •I can help you with preparing your dataset.
Please DM me with any questions you have.
Tomecki
in reply to Musharraf • • •Musharraf
in reply to Tomecki • • •Simply put:
Sonata is an inference engine that can theoretically drive any TTS model.
OptiSpeech is an actual model that generates speech, and the quality of the output depends on it.
Musharraf
in reply to Musharraf • • •Bill Dengler
in reply to Musharraf • • •Musharraf
in reply to Bill Dengler • • •Will DM you soon.
Luis Carlos
in reply to Musharraf • • •Musharraf
in reply to Luis Carlos • • •Yep. It is intended to be integrated into Sonata, which already runs on Windows/Android/iOS.
Luis Carlos
in reply to Musharraf • • •Musharraf
in reply to Luis Carlos • • •@luiscarlosgonzalez @NVAccess
I' afraid I cannot. The whole point of this and Sonata is to create high-quality, but very efficient and lightweight neural TTS. Coqui-TTS models are neither efficient nor lightweight.
Luis Carlos
in reply to Musharraf • • •Musharraf
in reply to Luis Carlos • • •Tortoise TTS is too heavy for a high-end server, let alone a standard computer or a mobile device.
This system is designed specifically for running on a standard CPU.
Scott
in reply to Musharraf • • •Musharraf
in reply to Scott • • •Sensitive content
Approx $50 for a one month of Google Colab subscription.
Scott
in reply to Musharraf • • •Sensitive content
Roberto Perez
in reply to Musharraf • • •Sensitive content
Musharraf
in reply to Roberto Perez • • •Thanks for your contribution. Really appreciate it!
For this month, Pneuma Solutions provided GPU resources for training. But I still accept donations to ensure we have enough resources for future training needs.
I'm happy to receive your donations via my colleague's paypal address:
paypal.me/geotts
or
Email:
info@geotts.ge
Please mention mush42/tts in the transaction note.
Zaplaťte uživateli Beka Gozalishvili pomocí služby PayPal.Me
PayPal.MeBen Blatchford
in reply to Musharraf • • •