🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

Here's a sneak peek of a crowd-based competition on sounding video generation starting from Oct. 1st! #ECCV2024 aicrowd.com/challenges/sou…

A list of diffusion works & tutorial (at #ISMIR2024) from our lab! [ML] arxiv.org/abs/2405.17251 #NeurIPS24 (GenWarp: Novel View Synthesis) arxiv.org/abs/2405.14822 #NeurIPS24 (PaGoDA: Multi-Scale 1 Step Generator) arxiv.org/abs/2310.02279 #ICLR24 (CTM: Fast Image Gen.)

🎶Large music models from our team🎶: 1. SoniDo🎼 for music mixing, demixing, transcription, etc. pdf: arxiv.org/abs/2411.01135 2. OpenMU🧙‍♂️ for music captioning, reasoning, lyric understanding, etc. pdf: arxiv.org/abs/2410.15573 code: mzhaojp22.github.io/open_music_und… demo ISMIR Conference :

I'm very happy to see that MMAudio, which my talented colleagues (mi141, A. Hayakawa, Takashi Shibuya) and intern (Rex Cheng) at Sony AI have invested their time and effort into, is being tested by so many creative people in X arXiv: arxiv.org/abs/2412.15322

When you register for #ICASSP2025, don't forget to select our tutorial on diffusion models for audio: 🎶Transforming Chaos into Harmony: Diffusion Models in Audio Signal Processing🎶 See you in Hyderabad, India!🇮🇳

Fast sampling methods for discrete diffusion from our lab 🏎️ Our sampling schedule optimization method (Jump Your Steps) is accepted at #ICLR2025 Another (Di4C) is about distllation for discrete diffusion!

We got 3 at #CVPR2025 and 6 at #ICLR2025. Kudos to my team members and interns for enormous effort to achieve this🪄 Paper details will be released soon✍️

#CVPR2025 Accepted Papers cvpr.thecvf.com/Conferences/20…

Here is MMAudio from Sony AI, one of our three papers accepted at #CVPR2025

The schedule of our #ICLR2025 papers is available. Don't miss poster sessions 1, 3, and 5, held on April 23rd, 24th, and 25th, respectively

🇮🇳Our tutorial on diffusion for audio signal processing was a great success, thanks to the amazing audience!! We're so grateful you chose our tutorial. We'll be uploading our materials soon!📃 sites.google.com/view/diffusion… #ICASSP2025

The #ICLR2025 week is coming soon! Check out our intern info as well as 6 papers from our lab: [Intern Positions] - Intern for Audio-Visual & Motion: ai.sony/joinus/job-rol… - Intern for Audio-Visual: sonyglobal.wd1.myworkdayjobs.com/en-US/SonyJapa… - Intern for 3D: ai.sony/joinus/job-rol… [Main

Our lab scored 3 papers at #ICML2025!🎉 1. Distillation of Discrete Diffusion through Dimensional Correlations (Di4C), joint work with いんそうさん x.com/kurowassann621… 2. Supervised Contrastive Learning from Weakly-labeled Audio Segments for Musical Version Matching (CLEWS)

It’s fulfilling that our lab’s audio papers were accepted to conferences with high H5-index rankings, where audio research is less common. I’d love to see more audio research highlighted there: - SoundCTM #ICLR2025 - MMDisCo #ICLR2025 - MMAudio #CVPR2025 - CLEWS #ICML2025

Gemini Diffusion looks great! Our recent work on optimizing sampling schedules (JYS, #ICLR2025) and a distillation method (Di4C, #ICML2025) for discrete diffusion introduces approaches that could potentially make dLLM inference even faster. If you're looking for literature on

MMDisCo at #ICLR2025 — a framework for co-generation across any modality combination, assuming modality-specific pretrained models. While "Seeing & Hearing" achieves multimodal alignment via ImageBind, MMDisCo uses a discriminator trained on synced/unsynced pairs as