[2025-11-06] 📚 Updates in #VGen
(1) ThinkSound: Chain-of-Thought Reasoning in Multimodal Large Language Models for Audio Generation and Editing
(2) Stable Part Diffusion 4D: Multi-View RGB and Kinematic Parts Video Generation
🔍 More at researchtrend.ai/communities/VGen
(1) ThinkSound: Chain-of-Thought Reasoning in Multimodal Large Language Models for Audio Generation and Editing
(2) Stable Part Diffusion 4D: Multi-View RGB and Kinematic Parts Video Generation
🔍 More at researchtrend.ai/communities/VGen
November 6, 2025 at 3:09 AM
🇨🇳ALIBABA INTRODUCES THINKSOUND, AN AI MODEL FOR INTERACTIVE, STEP-BY-STEP AUDIO GENERATION AND EDITING FOR VIDEO CONTENT ALIBABA UP OVER 3% IN HK #CHINA #AI #ALIBABA $BABA
July 16, 2025 at 2:15 AM
🇨🇳ALIBABA INTRODUCES THINKSOUND, AN AI MODEL FOR INTERACTIVE, STEP-BY-STEP AUDIO GENERATION AND EDITING FOR VIDEO CONTENT ALIBABA UP OVER 3% IN HK #CHINA #AI #ALIBABA $BABA
Alibaba: Launches ThinkSound, An AI Tool For Creating And Editing Audio Step-By-Step For Videos 🎧🎥
July 16, 2025 at 2:03 AM
Alibaba: Launches ThinkSound, An AI Tool For Creating And Editing Audio Step-By-Step For Videos 🎧🎥
Alibaba Group has unveiled ThinkSound, a new AI model designed for interactive, step-by-step audio generation and editing tailored for video content creation.
July 16, 2025 at 1:50 AM
Alibaba Group has unveiled ThinkSound, a new AI model designed for interactive, step-by-step audio generation and editing tailored for video content creation.
project page: ThinkSound-Demo.github.io
code: github.com/FunAudioLLM...
demo: huggingface.co/spaces/FunA...
code: github.com/FunAudioLLM...
demo: huggingface.co/spaces/FunA...
ThinkSound - a Hugging Face Space by FunAudioLLM
huggingface.co
July 11, 2025 at 4:14 PM
project page: ThinkSound-Demo.github.io
code: github.com/FunAudioLLM...
demo: huggingface.co/spaces/FunA...
code: github.com/FunAudioLLM...
demo: huggingface.co/spaces/FunA...
video-to-sound has been solved
there is a new spiritual successor to mmaudio called ThinkSound that supports chain-of-thought prompts for extremely accurate video-to-audio generation
kinda blown away:
there is a new spiritual successor to mmaudio called ThinkSound that supports chain-of-thought prompts for extremely accurate video-to-audio generation
kinda blown away:
July 11, 2025 at 4:14 PM
video-to-sound has been solved
there is a new spiritual successor to mmaudio called ThinkSound that supports chain-of-thought prompts for extremely accurate video-to-audio generation
kinda blown away:
there is a new spiritual successor to mmaudio called ThinkSound that supports chain-of-thought prompts for extremely accurate video-to-audio generation
kinda blown away:
ThinkSound (www.uneiaparjour.fr/thinksound/) génère la bande son de vidéos en la synchronisant et en ajoutant directement au fichier original, avec ou sans instructions.
Export mp4 et partage url.
Gratuit, open source et illimité.
#uneIAparjour #IA #audio #vidéo
Export mp4 et partage url.
Gratuit, open source et illimité.
#uneIAparjour #IA #audio #vidéo
ThinkSound - Une IA par jour
ThinkSound (démo : https://huggingface.co/spaces/FunAudioLLM/ThinkSound et documentation : https://github.com/FunAudioLLM/ThinkSound) génère la bande son de vidéos en la synchronisant et en ajoutant d...
www.uneiaparjour.fr
July 10, 2025 at 9:56 PM
ThinkSound (www.uneiaparjour.fr/thinksound/) génère la bande son de vidéos en la synchronisant et en ajoutant directement au fichier original, avec ou sans instructions.
Export mp4 et partage url.
Gratuit, open source et illimité.
#uneIAparjour #IA #audio #vidéo
Export mp4 et partage url.
Gratuit, open source et illimité.
#uneIAparjour #IA #audio #vidéo
Size matters! The large ThinkSound model (1.3B parameters) delivers the best results, but even smaller versions hold their own. The takeaway: more capacity means deeper reasoning and better audio, but the tech scales for different needs.
Paper: arxiv.org/abs/2506.21448
Paper: arxiv.org/abs/2506.21448
ThinkSound: Chain-of-Thought Reasoning in Multimodal Large Language Models for Audio Generation and Editing
While end-to-end video-to-audio generation has greatly improved, producing high-fidelity audio that authentically captures the nuances of visual content remains challenging. Like professionals in the ...
arxiv.org
July 5, 2025 at 1:22 AM
Size matters! The large ThinkSound model (1.3B parameters) delivers the best results, but even smaller versions hold their own. The takeaway: more capacity means deeper reasoning and better audio, but the tech scales for different needs.
Paper: arxiv.org/abs/2506.21448
Paper: arxiv.org/abs/2506.21448
ThinkSound uses a “gated fusion” mechanism to blend video and audio features. This dynamic approach lets the model adapt to each scene, improving synchronization and making the generated audio feel natural—no more awkward lags or mismatches.
July 5, 2025 at 1:22 AM
ThinkSound uses a “gated fusion” mechanism to blend video and audio features. This dynamic approach lets the model adapt to each scene, improving synchronization and making the generated audio feel natural—no more awkward lags or mismatches.
It’s not just about video or text—ThinkSound fuses both. By combining CLIP’s contrastive features with T5’s contextual reasoning, it nails both the details and the big picture. The result? Audio that’s not only realistic, but perfectly fits the scene.
Github: github.com/FunAudioLLM/...
Github: github.com/FunAudioLLM/...
GitHub - FunAudioLLM/ThinkSound: PyTorch implementation of [ThinkSound], a unified framework for generating audio from any modality, guided by Chain-of-Thought (CoT) reasoning.
PyTorch implementation of [ThinkSound], a unified framework for generating audio from any modality, guided by Chain-of-Thought (CoT) reasoning. - FunAudioLLM/ThinkSound
github.com
July 5, 2025 at 1:22 AM
It’s not just about video or text—ThinkSound fuses both. By combining CLIP’s contrastive features with T5’s contextual reasoning, it nails both the details and the big picture. The result? Audio that’s not only realistic, but perfectly fits the scene.
Github: github.com/FunAudioLLM/...
Github: github.com/FunAudioLLM/...
Imagine your video scenes coming alive with sound that actually makes sense—no more generic noise!
Project: thinksound-project.github.io
Movie Gen + ThinkSound
Project: thinksound-project.github.io
Movie Gen + ThinkSound
July 5, 2025 at 1:22 AM
Imagine your video scenes coming alive with sound that actually makes sense—no more generic noise!
Project: thinksound-project.github.io
Movie Gen + ThinkSound
Project: thinksound-project.github.io
Movie Gen + ThinkSound
ThinkSound is the open-source challenger shaking up the video-to-audio generation game.
Unlike commercial giants, ThinkSound brings chain-of-thought reasoning to multimodal AI, letting it generate audio that’s not just synced, but semantically spot-on.
VEO3 + ThinkSound 🧵1/5
Unlike commercial giants, ThinkSound brings chain-of-thought reasoning to multimodal AI, letting it generate audio that’s not just synced, but semantically spot-on.
VEO3 + ThinkSound 🧵1/5
July 5, 2025 at 1:22 AM
ThinkSound is the open-source challenger shaking up the video-to-audio generation game.
Unlike commercial giants, ThinkSound brings chain-of-thought reasoning to multimodal AI, letting it generate audio that’s not just synced, but semantically spot-on.
VEO3 + ThinkSound 🧵1/5
Unlike commercial giants, ThinkSound brings chain-of-thought reasoning to multimodal AI, letting it generate audio that’s not just synced, but semantically spot-on.
VEO3 + ThinkSound 🧵1/5
🎵 #Alibaba just dropped #ThinkSound - the first AI audio model that actually "thinks" before generating sound! Using Chain of Thought reasoning, it creates perfectly synced audio for every video frame, from baby cries to train sounds
aidisruption.ai/p/alibaba-op...
aidisruption.ai/p/alibaba-op...
Alibaba Open-Sources First CoT Audio Model, Mastering Audio-Visual Sync
Alibaba's ThinkSound: First CoT audio model for video dubbing. Generates frame-perfect sound effects using Chain of Thought reasoning. Open-source on GitHub.
aidisruption.ai
July 1, 2025 at 1:40 PM
🎵 #Alibaba just dropped #ThinkSound - the first AI audio model that actually "thinks" before generating sound! Using Chain of Thought reasoning, it creates perfectly synced audio for every video frame, from baby cries to train sounds
aidisruption.ai/p/alibaba-op...
aidisruption.ai/p/alibaba-op...
ThinkSound uses Chain-of-Thought reasoning to enable stepwise, interactive audio generation and editing for videos, guided by a multimodal large language model and a unified audio foundation model; state-of-the-art performance on video-to-audio generation.
ThinkSound: Chain-of-Thought Reasoning in Multimodal Large Language Models for Audio Generation and Editing
Huadai Liu, Jialei Wang, Kaicheng Luo, Wen Wang, Qian Chen, Zhou Zhao, Wei Xue
arxiv.org
June 27, 2025 at 11:15 AM
ThinkSound uses Chain-of-Thought reasoning to enable stepwise, interactive audio generation and editing for videos, guided by a multimodal large language model and a unified audio foundation model; state-of-the-art performance on video-to-audio generation.
Huadai Liu, Jialei Wang, Kaicheng Luo, Wen Wang, Qian Chen, Zhou Zhao, Wei Xue: ThinkSound: Chain-of-Thought Reasoning in Multimodal Large Language Models for Audio Generation and Editing https://arxiv.org/abs/2506.21448 https://arxiv.org/pdf/2506.21448 https://arxiv.org/html/2506.21448
June 27, 2025 at 6:35 AM
Huadai Liu, Jialei Wang, Kaicheng Luo, Wen Wang, Qian Chen, Zhou Zhao, Wei Xue: ThinkSound: Chain-of-Thought Reasoning in Multimodal Large Language Models for Audio Generation and Editing https://arxiv.org/abs/2506.21448 https://arxiv.org/pdf/2506.21448 https://arxiv.org/html/2506.21448
June 1, 2025 at 12:39 PM
I had to listen to this album with my new headphones. 😯🥰 So much details. 😯😀♥️ #AlitaArmy #Headphones #AudioExperience #Wow #ThinkSound #AlitaSoundtrack
May 31, 2025 at 6:55 PM
I had to listen to this album with my new headphones. 😯🥰 So much details. 😯😀♥️ #AlitaArmy #Headphones #AudioExperience #Wow #ThinkSound #AlitaSoundtrack
Just shopped Canadian with the purchase of gorgeous headphones from Ontario-based thinksound.
April 29, 2025 at 1:07 AM
Just shopped Canadian with the purchase of gorgeous headphones from Ontario-based thinksound.
Thinksound moved from my former neck of the woods in NH to ON a ways back. Given the CEO's focus on supply chain efficiency I have to imagine it's working well there. Kanto also makes speakers in BC.
TON of capable grads in BC, ON. Tax incentives, too! www.investcanada.ca/programs-inc...
TON of capable grads in BC, ON. Tax incentives, too! www.investcanada.ca/programs-inc...
Programs and incentives | Invest in Canada
Canada’s investment programs and incentives make it easy for your company to expand and succeed. Learn more.
www.investcanada.ca
March 5, 2025 at 9:49 PM
Thinksound moved from my former neck of the woods in NH to ON a ways back. Given the CEO's focus on supply chain efficiency I have to imagine it's working well there. Kanto also makes speakers in BC.
TON of capable grads in BC, ON. Tax incentives, too! www.investcanada.ca/programs-inc...
TON of capable grads in BC, ON. Tax incentives, too! www.investcanada.ca/programs-inc...