Julian Lenz
@JLenzyy
Followers
633
Following
3K
Media
49
Statuses
523
Audio AI research engineer w/ Lemonaide. prev. Neutone, Okio. MSc in Audio Computation at UPF. Opinions are a biased interpolation of my training dataset.
Joined July 2022
I want to urge fellow audio-AI researchers to stop talking about 'solving' audio. People, we work in a creative domain You don't 'solve' art. Bach did not 'solve' baroque music. The term alienates us from musicians. We should be here to empower and enhance creativity!
9
40
249
Kurt Vonnegut on buying an envelope
order your groceries. saves an extra 1-2 hours a week. compounded, thats 50-100 hours a year. a whole week work of work you gain. i promise that time is more important then the $10 delivery fee
289
17K
116K
🥳 Announcing Stable Audio Open Small - Fastest text2audio evarrr?! - Introducing ARC: adversarial-relativistic-contrastive training for accelerating diffusion models - 75ms on H100 Weights https://t.co/azeQiDRU7q Blog https://t.co/guUFZ1muvq Paper
huggingface.co
6
36
149
Hacked together a real-time piano continuation demo this weekend. Piano sound = real-time input. Electric piano sound = real-time (ai) continuation streamed from my server over the LAN. Gotta get this hooked up to an acoustic self-playing piano, for both input/output, next week!
3
4
22
Tantalizing how much we don't understand
0
0
2
One of the deciding factors for me to join @lemonaidexmusic was seeing @mjxmusic (the CEO) just straight up jamming with Seeds constantly on IG. We are building stuff that we want to have fun playing with.
The surprising thing about tech leadership I've observed is just using the product automatically puts you in the top 10% of leaders. One of the craziest cheat codes.
0
0
8
New article feat. @lemonaidexmusic is out! We talk about the challenges and joys of developing the world's first @fairlytrained certified MIDI VST. Thank you @EzraSandzer for your phenomenal support 🙏
My new https://t.co/66fghOkUKr article about the AI MIDI generator @lemonaidexmusic just went live. It includes a deep dive interview with their founder @mjxmusic and lead developers @anirudh3 and @JLenzyy. You can read below 👇 https://t.co/Q0DqG63ahV
1
0
8
'oops'
lawyers for The New York Times and other outlets spent over 150 hours searching OpenAI’s data as part of a lawsuit. OpenAI accidentally deleted all of evidence that was hosted on virtual machines 😬
0
0
1
Probably this is a capitalism thing. "Hey, wouldn't it be cool if you could have books and music and drawings that are made for YOU?" But I wonder, will a society full of self-reflecting AI prompters silently decay? Will the incentives to be a profoundly weird
3
0
2
Something I dislike in many AI art convos is the implicit assumption that making art for yourself, as an amateur, is better than consuming art made by others. Art is powerful when it challenges, and most will not prompt an LLM to write a book that pushes against their priors.
3
3
12
🎶✨ Tired of manually specifying each note to mask during music inpainting? 🎵 We propose an approach for training a model without supervision that suggests where to make edits for better sounding music! 🎧 We call it RefinPaint for Music Proofreading. 🎨🖋️ While there are
1/2 🚀Last week at #ISMIR2024 in San Francisco, we presented #RefinPaint: a collaborative music system with two agents—one generates music, and the other critiques to refine it. The Critic Agent, trained on synthetic data, unlocks unique interactions in automatic music generation
1
6
28
Incredible
I've combined some of these together into a new #neuro #visualization "In the eye of a fly": Using the @FlyWireNews visual system wiring diagram, #Brian2 spiking #neuron simulator, and #blender. Information flow through the chiasms, across a topographic map.
0
0
3
Brilliant keynote @ednewtonrex, and thank you for the shout out! One of the proudest moments I've had as a researcher. A room full of audio scientists applauding the companies that use fairly licensed data.
0
3
33
Thrilled to be in San Francisco for #ISMIR2024 repping @lemonaidexmusic. We will present our paper on tokenization + generation of expressive MIDI. Come chat about all things symbolic!
🎵🪇 Expressive Generation and Variations on Symbolic Music 🎶🎻 So excited to announce that the first 🍋@lemonaidexmusic research paper, which received a 'strong accept' at ISMIR is now out on Arxiv! Link below, and more details in this short 🧵 https://t.co/HjN2brTfqd
2
2
14
I've consolidated my initial research on data center freshwater consumption and combined it with some broad strokes about the merits of MIDI generation. TLDR; The small file size of MIDI makes it cheaper to model, generate, and hold in storage. https://t.co/SLZe98WwMv
midi.org
The ethics of AI music became a heated topic at industry panels in 2024, sparking debates around the notion of “fair use”. AI music tech companies have admitted to training [...]
1
4
17
Would be extra fun in conjunction with more standard VAE tricks. Pick a single latent vector, and stay there, listening to how it morphs as you gradually pull the model into different training sets, and destroy the weights with a negative loss function
1
0
3
late-night musings: would be neat to have a live(ish) audio generation model that is being trained / weights updated during the performance. A form of 'controllability' would be steering it towards different training data, modifying the learning rate, and even 'untraining' it
4
0
11
lego star wars has no right to be as fun and hilarious as it is
0
0
1
AN ATTIC RED-FIGURED FIGURAL MUG Circa 430-420 B.C. 7 7/8 in. (20 cm) high (Christie's Auctions)
0
16
76
🚨🎶ISMIR 2024 Final Paper Sessions!🎶🚨 The final paper sessions have been announced! Each session's first paper is a best paper award nominee. Congratulations to all the nominees & to all authors at ISMIR 2024! Please visit https://t.co/FtAjpc3kM8
#ismir2024 #paperpresentation
0
2
16