ML Research and Engineering @ Disney Software developer PhD w/ Adobe & UoSurrey I make stuff from time to time. (eg xVASynth) https://teapaw.org/t.co/rAiFoZ5STN
Joined February 2013
- Tweets52
- Following254
- Followers140
- Likes131
Dan Ruta (@danruta.bsky.social) retweeted
The 𝐱𝐕𝐀𝐒𝐲𝐧𝐭𝐡 𝐓𝐓𝐒 app is on Spaces!
😎Thanks to @Pendrokar ! Link in tweet below!
Link to the 🤗 Space:
huggingface.co/spaces/Pendro…
Dan Ruta (@danruta.bsky.social) retweeted
Have you explored the 𝐱𝐕𝐀𝐒𝐲𝐧𝐭𝐡 𝐓𝐓𝐒 app on Spaces yet?
😎The app is a voice actor oriented TTS, mostly used by mods of The Elder Scrolls games series.🤠App synthesizes text in a few seconds even on CPUs
Combined with STT & LLMs, Project helps in 'speaking with NPCs'😄
Following my Viva pass last week, I'm excited to share the recording of my pre-viva open presentation, where I talk through the 6 papers written during my PhD, covering multi-modal artistic style representation learning, and generative art synthesis.
youtube.com/watch?v=msoQI9EX…
1
224
Better late than never!!
The code and weights for the ALADIN artistic style representation model (the ViT variant evolved in StyleBabel) have now been cleared for public open source release!
Code:
github.com/DanRuta/ALADIN
Papers:
arxiv.org/abs/2103.09776
arxiv.org/abs/2203.05321
1
278
xVASynth v3.0 is now out!
Now supporting v3 models which can do 28 languages, emotion control, custom styles, voice conversion, voice crafting and more, in addition to the pitch/energy/duration control. Trainable via xVATrainer v1.2
youtube.com/watch?v=5u4xpI-c…
Just released xVATrainer, a tool that anyone can use to create AI text-to-speech models, without any Machine Learning or programming experience.
Open source and available on Steam/@NexusSites
youtube.com/watch?v=PXv_SeTW…
Dan Ruta (@danruta.bsky.social) retweeted
StyleBabel: Artistic Style Tagging and Captioning
abs: arxiv.org/abs/2203.05321
xVASynth v2.1 is now out on Steam!
store.steampowered.com/app/1…
Check out the v2 showcase video ICYMI.
youtube.com/watch?v=W-9SFoNu…
I'm excited to announce this was accepted to @ICCV_2021 🥳
I'm happy to share some of our latest work in our research paper:
ALADIN: All Layer Adaptive Instance Normalization for Fine-grained Style Similarity
where we learn a state-of-the-art representation of artistic style in digital artistic media.
arxiv.org/abs/2103.09776
1
3
Dan Ruta (@danruta.bsky.social) retweeted
The AI revolution coming to Skyrim and The Witcher 3’s mod scenes
pcgamesn.com/the-elder-scrol…
I'm happy to share some of our latest work in our research paper:
ALADIN: All Layer Adaptive Instance Normalization for Fine-grained Style Similarity
where we learn a state-of-the-art representation of artistic style in digital artistic media.
arxiv.org/abs/2103.09776
Dan Ruta (@danruta.bsky.social) retweeted
one top @behance request has long been a hyper-intelligent IMAGE search functionality (not just projects). it's now live, with an ai-powered "more like this" hover state for every image. 🔥 behance.net/search/images
The AI future of voice acting (maybe):
Released v1.0 of xVASynth today, an app using neural speech synthesis to generate controllable voice acting lines in 34+ voices across 53+ characters, with plenty more coming.
youtube.com/watch?v=xqJ-ujWv…
#MachineLearning
Dan Ruta (@danruta.bsky.social) retweeted
A very warm welcome to the new members in CVSSP. A record 33 people are joining us today (23 PhDs, 5 Researchers and 5 Visitors). 🎉🎊🍾😁😊
Great article if you love Game of Thrones, and you're interested in learning Prolog. Even if you don't! “How to learn Prolog by watching Game of Thrones” by Rachel Wiles link.medium.com/Cvn0XRjVNV
#GameofThrones
Been playing around with machine learning based voice synthesis, and made xVASynth, an app which lets you generate speech in the style of specific characters from Bethesda games (Skyrim,Fallout,...). It's early days, but it sort of works:
youtube.com/watch?v=ofN0yMI9…
#MachineLearning
3
If you're looking to learn some #WebAssembly, pick up @codelikeawolf 's new book, "Learn WebAssembly". I had the honour of being a technical reviewer for it, so I already know it's got a bunch of good stuff in there.
Just posted an article on the work me, Louis and Tom did in our paper, earlier in the year.
medium.freecodecamp.org/how-…
A colourful, animated terminal histogram library, because apparently my procrastination knows no bounds: github.com/DanRuta/CMD-Chart #javascript
Released v3.3.0 of jsNet, with terminal/console confusion matrix, mnist dev environment example, volume input, and a couple of configs for faster training.
github.com/DanRuta/jsNet
#ai #machinelearning
2