正在加载视频...

视频加载失败

加载此视频时出现问题。这可能是由于临时网络问题，或视频可能不可用。

DeepMind just dropped Lyria, a state-of-the-art music generative model. Last year, I said 2022 was the year of pixels and 2023 would be the year of soundwaves. We are making great progress here! The most impressive demo is converting humming to a full instrument suite. I think Lyria will... unlock all the operators we are used to in image models: text-based editing, style transfer, in-painting (fill out tracks), out-painting (continue a track), super-resolution, etc. Lyria is deployed as an intuitive software tool to musicians, in partnership with YouTube. This is the right move: ship the model! With enough artists on board, Lyria could spin a data flywheel that learns from the artists' feedback and editing signals. Congrats to DeepMind for shipping! Launch blog:show more

Jim Fan

384,982 subscribers

384,184 次观看 • 2 年前 •via X (Twitter)

教育科学技术音乐

Anya Rossi• Live Now

Private livecam show

10 条评论

René Schulte 的头像

René Schulte2 年前

Honestly it gets a bit boring with all the announcements of things you can not actually use or have to sign-up for a wait list... Don't get me wrong, Lyria sounds like the most impressive to date and I get there need to be guardrails but as content creator I need to try this stuff myself.

Simon ➠ e/acc 的头像

Simon ➠ e/acc2 年前

Bard would have been a better name🙄

Per-Anders Edwards 的头像

Per-Anders Edwards2 年前

What a miserable empty future we’re moving towards. There will never be another shared experience, another exciting opportunity to look forward to. Anything fresh ever again. Anything discovered. Just a world turned into a sea of vapid meaningless personalized and lonely “content”. Stultification. If you’re that shallow you either are excited at the potential for your own greed, but don’t yet realize what you have will be worth nothing, or that you imagine that the current trends in AI are big wow’s, then you’re already a victim of incredibly low standards, and just want to drag the rest of us down with you. Equality in the gutter, never the stars.

Evan Knight 的头像

Evan Knight2 年前

Riffusion is available right now @riffusionai

Darryl Mason 的头像

Darryl Mason2 年前

I intend to spend the first month after I get access making symphonic music from the bird songs of the magpies and lorikeets in my yard. Can't wait!

Gonzalo Espinoza Graham 🏴‍☠️ 的头像

Gonzalo Espinoza Graham 🏴‍☠️2 年前

They shipped a blog post

Will 的头像

Will2 年前

on the one hand, wow! amazing technical accomplishment, and, yes, it will help people flesh out ideas faster. on the other hand, wow, this will help tasteless people create trash so fast. Pop music is already painful enough!

Aleksa Gordić 🍿🤖 的头像

Aleksa Gordić 🍿🤖2 年前

i stopped caring about announcements where a company is shipping a blog post unless the blog is insightful/interesting as for example one of your prev posts on how you trained on all internal docs

dt 🐙 的头像

dt 🐙2 年前

wen access

Daniel 的头像

Daniel2 年前

take a look at much better

相关视频

Last month, we released Lyria 3, enabling you to create tracks with lyrics from text, image, or video prompts. Now, we’re introducing Lyria 3 Pro, which expands upon our music generation model to offer additional advanced capabilities. What’s really special about this upgrade is that the model now understands the architecture of music. This makes it possible to prompt for intros, verses, choruses and bridges + generate songs with more complex transitions. You can also create tracks up to 3 minutes long, a big change from previous models that were limited to 30 second tracks. Use Lyria 3 Pro to build upon your existing creativity. We’re excited for your beats to drop 🎶

Last month, we released Lyria 3, enabling you to create tracks with lyrics from text, image, or video prompts. Now, we’re introducing Lyria 3 Pro, which expands upon our music generation model to offer additional advanced capabilities. What’s really special about this upgrade is that the model now understands the architecture of music. This makes it possible to prompt for intros, verses, choruses and bridges + generate songs with more complex transitions. You can also create tracks up to 3 minutes long, a big change from previous models that were limited to 30 second tracks. Use Lyria 3 Pro to build upon your existing creativity. We’re excited for your beats to drop 🎶

Google AI

81,583 次观看 • 5 个月前

Lyria 3 came out from google today, so naturally I had to use it for generative lullabies. now, when the monitor detects the baby is crying, it can play a 30 sec lullaby pre-generated to the specific baby's tastes (based on what's worked historically) using Lyria 3 Clip. this demo showcases this generative lullaby capability, whereas the last showcased playing soothing phrases in the mother's cloned voice to calm the baby. this is using the OpenHome devkit as hardware. more to come!

Lyria 3 came out from google today, so naturally I had to use it for generative lullabies. now, when the monitor detects the baby is crying, it can play a 30 sec lullaby pre-generated to the specific baby's tastes (based on what's worked historically) using Lyria 3 Clip. this demo showcases this generative lullaby capability, whereas the last showcased playing soothing phrases in the mother's cloned voice to calm the baby. this is using the OpenHome devkit as hardware. more to come!

shira

110,653 次观看 • 5 个月前

TokenFlow: Consistent Diffusion Features for Consistent Video Editing paper page: The generative AI revolution has recently expanded to videos. Nevertheless, current state-of-the-art video models are still lagging behind image models in terms of visual quality and user control over the generated content. In this work, we present a framework that harnesses the power of a text-to-image diffusion model for the task of text-driven video editing. Specifically, given a source video and a target text-prompt, our method generates a high-quality video that adheres to the target text, while preserving the spatial layout and motion of the input video. Our method is based on a key observation that consistency in the edited video can be obtained by enforcing consistency in the diffusion feature space. We achieve this by explicitly propagating diffusion features based on inter-frame correspondences, readily available in the model. Thus, our framework does not require any training or fine-tuning, and can work in conjunction with any off-the-shelf text-to-image editing method. We demonstrate state-of-the-art editing results on a variety of real-world videos.

TokenFlow: Consistent Diffusion Features for Consistent Video Editing paper page: The generative AI revolution has recently expanded to videos. Nevertheless, current state-of-the-art video models are still lagging behind image models in terms of visual quality and user control over the generated content. In this work, we present a framework that harnesses the power of a text-to-image diffusion model for the task of text-driven video editing. Specifically, given a source video and a target text-prompt, our method generates a high-quality video that adheres to the target text, while preserving the spatial layout and motion of the input video. Our method is based on a key observation that consistency in the edited video can be obtained by enforcing consistency in the diffusion feature space. We achieve this by explicitly propagating diffusion features based on inter-frame correspondences, readily available in the model. Thus, our framework does not require any training or fine-tuning, and can work in conjunction with any off-the-shelf text-to-image editing method. We demonstrate state-of-the-art editing results on a variety of real-world videos.

AK

143,469 次观看 • 3 年前

Yes, jellyfish DJ’d the “opening set” at #GoogleIO this year. 🪼Let’s dive in 🌊 We partnered with Monterey Bay Aquarium to build a generative music experiment where moon jellies, live-streamed from their home at the aquarium, controlled music made by Lyria 3. The jellies were tracked by a lightweight object detection model running on a tiny Coral NPU housed on a Synaptics Coralboard — the same type of object detection model used to power conservation efforts in the Great Barrier Reef. The Coralboard sent the jellies’ coordinates to our music generation engine. Each quadrant of the screen represented an instrument — drums, bass, chords or melody. The number of jellies in each quadrant determined the weighting of those sounds, translating their movements into music.

Yes, jellyfish DJ’d the “opening set” at #GoogleIO this year. 🪼Let’s dive in 🌊 We partnered with Monterey Bay Aquarium to build a generative music experiment where moon jellies, live-streamed from their home at the aquarium, controlled music made by Lyria 3. The jellies were tracked by a lightweight object detection model running on a tiny Coral NPU housed on a Synaptics Coralboard — the same type of object detection model used to power conservation efforts in the Great Barrier Reef. The Coralboard sent the jellies’ coordinates to our music generation engine. Each quadrant of the screen represented an instrument — drums, bass, chords or melody. The number of jellies in each quadrant determined the weighting of those sounds, translating their movements into music.

Google

65,696 次观看 • 2 个月前

EOY Recap: The Daily Program, Season II In 2025, the program continued as a space for artistic experimentation around AI video. This year, we invited 70 artists to present work that extends the survey we began in 2023. AI video remains an unstable genre because the tools change quickly and unevenly, with each new release altering what motion and continuity do to storytelling or formal experimentation. That instability is not only a constraint. It is also a condition that invites risk, iteration, and the proposal of new visual grammars. Across the year, artists worked with a wide range of approaches and model families, from GAN-based systems to Sora, alongside other contemporary text-to-video and image-to-video tools. ↓ Below is a longer reflection on why we think this initiative matters

EOY Recap: The Daily Program, Season II In 2025, the program continued as a space for artistic experimentation around AI video. This year, we invited 70 artists to present work that extends the survey we began in 2023. AI video remains an unstable genre because the tools change quickly and unevenly, with each new release altering what motion and continuity do to storytelling or formal experimentation. That instability is not only a constraint. It is also a condition that invites risk, iteration, and the proposal of new visual grammars. Across the year, artists worked with a wide range of approaches and model families, from GAN-based systems to Sora, alongside other contemporary text-to-video and image-to-video tools. ↓ Below is a longer reflection on why we think this initiative matters

Fellowship

18,627 次观看 • 8 个月前

co-seeing art with machine: an interactive demo where i browse the visual responses of a multimodal llm to prompts about the painting, from “who’s in here?” and to “what’s unusual about this painting?” the model replies not just with words, but with visual annotations as well.

co-seeing art with machine: an interactive demo where i browse the visual responses of a multimodal llm to prompts about the painting, from “who’s in here?” and to “what’s unusual about this painting?” the model replies not just with words, but with visual annotations as well.

Kat ⊷ the Poet Engineer

37,359 次观看 • 1 年前

TurboEdit Instant text-based image editing discuss: We address the challenges of precise image inversion and disentangled image editing in the context of few-step diffusion models. We introduce an encoder based iterative inversion technique. The inversion network is conditioned on the input image and the reconstructed image from the previous step, allowing for correction of the next reconstruction towards the input image. We demonstrate that disentangled controls can be easily achieved in the few-step diffusion model by conditioning on an (automatically generated) detailed text prompt. To manipulate the inverted image, we freeze the noise maps and modify one attribute in the text prompt (either manually or via instruction based editing driven by an LLM), resulting in the generation of a new image similar to the input image with only one attribute changed. It can further control the editing strength and accept instructive text prompt. Our approach facilitates realistic text-guided image edits in real-time, requiring only 8 number of functional evaluations (NFEs) in inversion (one-time cost) and 4 NFEs per edit. Our method is not only fast, but also significantly outperforms state-of-the-art multi-step diffusion editing techniques.

TurboEdit Instant text-based image editing discuss: We address the challenges of precise image inversion and disentangled image editing in the context of few-step diffusion models. We introduce an encoder based iterative inversion technique. The inversion network is conditioned on the input image and the reconstructed image from the previous step, allowing for correction of the next reconstruction towards the input image. We demonstrate that disentangled controls can be easily achieved in the few-step diffusion model by conditioning on an (automatically generated) detailed text prompt. To manipulate the inverted image, we freeze the noise maps and modify one attribute in the text prompt (either manually or via instruction based editing driven by an LLM), resulting in the generation of a new image similar to the input image with only one attribute changed. It can further control the editing strength and accept instructive text prompt. Our approach facilitates realistic text-guided image edits in real-time, requiring only 8 number of functional evaluations (NFEs) in inversion (one-time cost) and 4 NFEs per edit. Our method is not only fast, but also significantly outperforms state-of-the-art multi-step diffusion editing techniques.

AK

16,062 次观看 • 2 年前

Exciting news! MegaSAM code is out🔥 & the updated Shape of Motion results with MegaSAM are really impressive! A year ago I didn't think we could make any progress on these videos: Huge congrats to everyone involved and the community 🎉

Exciting news! MegaSAM code is out🔥 & the updated Shape of Motion results with MegaSAM are really impressive! A year ago I didn't think we could make any progress on these videos: Huge congrats to everyone involved and the community 🎉

Angjoo Kanazawa

95,540 次观看 • 1 年前

IT'S SO OVER I've been generating with Producer AI (powered by Lyria 3 from Google) for the last couple of hours, and I'm in shock. The music industry is done as we know it. It's going to be 95%+ AI music going forward. Listen to these songs I generated: 1. Aliens in Atlanta

IT'S SO OVER I've been generating with Producer AI (powered by Lyria 3 from Google) for the last couple of hours, and I'm in shock. The music industry is done as we know it. It's going to be 95%+ AI music going forward. Listen to these songs I generated: 1. Aliens in Atlanta

Alex Patrascu

13,481 次观看 • 6 个月前

Happy Wrapped! This is obviously a special moment for us and it’s still pretty wild to me how big it’s gotten year over year. In fact, after about the third year we did this, I specifically remember thinking that we had reached the peak and couldn’t go bigger. 10ish years later I was obviously very wrong. Here are some fun highlights over the last decade on how Wrapped has evolved. Thanks to all the fans, artists and creators for making this year what it was. And congrats to the hundreds of Spotifiers who work year-round to bring Wrapped to life.

Happy Wrapped! This is obviously a special moment for us and it’s still pretty wild to me how big it’s gotten year over year. In fact, after about the third year we did this, I specifically remember thinking that we had reached the peak and couldn’t go bigger. 10ish years later I was obviously very wrong. Here are some fun highlights over the last decade on how Wrapped has evolved. Thanks to all the fans, artists and creators for making this year what it was. And congrats to the hundreds of Spotifiers who work year-round to bring Wrapped to life.

Daniel Ek

22,780 次观看 • 1 年前

The Niceaunties AI Video Grant At Fellowship, we believe in creating an environment for experimentation and the development of ideas. That is why, when niceaunties proposed a grant program for mentorship and the creation of AI videos, we immediately said yes. The program will help support six artists, to create a Solo Show for the Fellowship Daily program. The commitment is to produce 8 new videos based on a theme proposed by Niceaunties. This will be an open call for all the Daily Artists and the AI video community at large. Everyone is welcome to reply/submit a video for consideration. Each grantee will receive a total of 2ETH provided by Niceaunties and Fellowship, as well as mentorship from Niceaunties and alejandro cartagena. We encourage artists applying to consider this grant as a space of critical feedback and of experimentation. Artists are encouraged to be open to new ideas and to explore alternative ways of how to think of their practice. These funds and curatorial support are there to open up time for the creators to engage and create a space ideal for critical feedback. We believe that developing an artistic voice comes from experimentation and confrontation of ideas. All 8 pieces produced during the month will be auctioned on the daily . xyz site as a one-day Solo Show. All proceeds of the sales will go to the artists. We are looking for creators who want to expand their ideas and take their art to the next level. We thank Niceaunties for their passion and commitment to AI art and the art community. This is the open call post. Please follow, retweet and reply/submit a video for consideration. You have until Friday the 5th of July to submit the work. 📅 The theme for this month is CONNECTIONS. Good luck everyone!

The Niceaunties AI Video Grant At Fellowship, we believe in creating an environment for experimentation and the development of ideas. That is why, when niceaunties proposed a grant program for mentorship and the creation of AI videos, we immediately said yes. The program will help support six artists, to create a Solo Show for the Fellowship Daily program. The commitment is to produce 8 new videos based on a theme proposed by Niceaunties. This will be an open call for all the Daily Artists and the AI video community at large. Everyone is welcome to reply/submit a video for consideration. Each grantee will receive a total of 2ETH provided by Niceaunties and Fellowship, as well as mentorship from Niceaunties and alejandro cartagena. We encourage artists applying to consider this grant as a space of critical feedback and of experimentation. Artists are encouraged to be open to new ideas and to explore alternative ways of how to think of their practice. These funds and curatorial support are there to open up time for the creators to engage and create a space ideal for critical feedback. We believe that developing an artistic voice comes from experimentation and confrontation of ideas. All 8 pieces produced during the month will be auctioned on the daily . xyz site as a one-day Solo Show. All proceeds of the sales will go to the artists. We are looking for creators who want to expand their ideas and take their art to the next level. We thank Niceaunties for their passion and commitment to AI art and the art community. This is the open call post. Please follow, retweet and reply/submit a video for consideration. You have until Friday the 5th of July to submit the work. 📅 The theme for this month is CONNECTIONS. Good luck everyone!

The Daily Program

18,353 次观看 • 2 年前

Today, we're open sourcing the code behind "The Infinite Crate," our VST/AU plugin that lets you play with Lyria RealTime directly inside your favorite DAW! 🎧🎶 Since we released The Infinite Crate last year, it’s been used by some of our favorite artists - including a wonderful showcase with Daito Manabe in Tokyo - and being featured as a top new music tool at NAMM 2026. We’re now fully open sourcing the plugin for developers to fork, modify, and make their own under the permissive Apache 2.0 license. 💾 Get it here:

Today, we're open sourcing the code behind "The Infinite Crate," our VST/AU plugin that lets you play with Lyria RealTime directly inside your favorite DAW! 🎧🎶 Since we released The Infinite Crate last year, it’s been used by some of our favorite artists - including a wonderful showcase with Daito Manabe in Tokyo - and being featured as a top new music tool at NAMM 2026. We’re now fully open sourcing the plugin for developers to fork, modify, and make their own under the permissive Apache 2.0 license. 💾 Get it here:

Jesse Engel

20,861 次观看 • 5 个月前

Ask and you shall receive! This is a work in progress project that I worked on yesterday and it's a little bit out of my comfort zone, but I was just trying to practice this style of editing. Any thoughts and feedback is very much appreciated! Credits are in the comments.

Ask and you shall receive! This is a work in progress project that I worked on yesterday and it's a little bit out of my comfort zone, but I was just trying to practice this style of editing. Any thoughts and feedback is very much appreciated! Credits are in the comments.

AlexNSFW 💙 (COMMS OPEN)

20,024 次观看 • 2 年前

Bird SQL is an impressive new tool, based on language models, for searching Twitter. Tools like this are changing the way we interact with information. If used in the right way, signals are becoming easier to find through the use of LLMs. Try it here:

Bird SQL is an impressive new tool, based on language models, for searching Twitter. Tools like this are changing the way we interact with information. If used in the right way, signals are becoming easier to find through the use of LLMs. Try it here:

elvis

95,883 次观看 • 3 年前

“At Turning Point Action, we wanted to put on an event, a training that is for you, the citizen, that is for We the People, that is for a restoration of this Republic. In the next couple of days, you will be challenged. You will be encouraged... This is going to be the end of spectator politics... Instead, we are going to make a commitment that this year, this calendar year, will be the hardest year we have ever worked.” #PEOPLES2024

“At Turning Point Action, we wanted to put on an event, a training that is for you, the citizen, that is for We the People, that is for a restoration of this Republic. In the next couple of days, you will be challenged. You will be encouraged... This is going to be the end of spectator politics... Instead, we are going to make a commitment that this year, this calendar year, will be the hardest year we have ever worked.” #PEOPLES2024

Charlie Kirk

344,272 次观看 • 2 年前

Here is DLSS5 working on an animation I made in Unreal. It uses the existing geometry, animation, textures and lighting to render. It doesn't make stuff up out of thin air. There is zero generative AI. This is the exact use of AI that people have wanted. As a tool where artists maintain full art direction. Neural rendering makes creating amazing art with real 3D models easier and more accessible. All 3D artists should be excited. Here is another explanation by Daniel Vávra ⚔ Is this a 3D model? I encourage you to create a 3D model or animation sometime and try it out for yourself, it can be fun creating something!

Here is DLSS5 working on an animation I made in Unreal. It uses the existing geometry, animation, textures and lighting to render. It doesn't make stuff up out of thin air. There is zero generative AI. This is the exact use of AI that people have wanted. As a tool where artists maintain full art direction. Neural rendering makes creating amazing art with real 3D models easier and more accessible. All 3D artists should be excited. Here is another explanation by Daniel Vávra ⚔ Is this a 3D model? I encourage you to create a 3D model or animation sometime and try it out for yourself, it can be fun creating something!

Unreal Sensei

119,013 次观看 • 1 天前

I got a diffusion model to run really fast, glued it to a LLM, and it blows my mind how much of the interactivity we expect from software (editing text, dragging things around, clicking on "links") just works™️ in pixel space. The future of software is bright

I got a diffusion model to run really fast, glued it to a LLM, and it blows my mind how much of the interactivity we expect from software (editing text, dragging things around, clicking on "links") just works™️ in pixel space. The future of software is bright

Zain Shah

22,649 次观看 • 1 个月前

Spike believes there is an urgency to win a Super Bowl for the Eagles this year. "I don't think it can be a transition year. I think it has to be a year where the only result that feels like a success is the Super Bowl because I just don't know how long this era of players and coaches will be here."

Spike believes there is an urgency to win a Super Bowl for the Eagles this year. "I don't think it can be a transition year. I think it has to be a year where the only result that feels like a success is the Super Bowl because I just don't know how long this era of players and coaches will be here."

SPORTSRADIO 94WIP

17,274 次观看 • 21 天前

What Evidence Could Prove Epstein is Alive? “So you take all of these now in totality, from the passport, to the conditions in the jail cell, the trust, the escape plan, the condition of the body when he is wheeled out on the gurney, and walk your way all the way down to post-life signals that carry some weight. The option with the fewest problems of the three is that he left that cell alive. We want proof of life. We started the reward pool and put in $40,000 of our own money, and people are contributing to this reward pool. If we’re wrong, and I’m willing to be wrong, in fact, I would prefer to be wrong. I take no satisfaction being right about this. But if we’re right, that reward pool will grow and grow year over year as a long-term commitment to justice and truth.”

What Evidence Could Prove Epstein is Alive? “So you take all of these now in totality, from the passport, to the conditions in the jail cell, the trust, the escape plan, the condition of the body when he is wheeled out on the gurney, and walk your way all the way down to post-life signals that carry some weight. The option with the fewest problems of the three is that he left that cell alive. We want proof of life. We started the reward pool and put in $40,000 of our own money, and people are contributing to this reward pool. If we’re wrong, and I’m willing to be wrong, in fact, I would prefer to be wrong. I take no satisfaction being right about this. But if we’re right, that reward pool will grow and grow year over year as a long-term commitment to justice and truth.”

Shawn Ryan

97,098 次观看 • 2 个月前

Hopping back on here from the other site to announce WE MAKE THE TV GLOW is coming in May to celebrate the 1 year anniversary of I Saw the TV Glow!! I would love to stack the night's programming with videos by trans and nb artists - if that is you, please reach out 💕

Hopping back on here from the other site to announce WE MAKE THE TV GLOW is coming in May to celebrate the 1 year anniversary of I Saw the TV Glow!! I would love to stack the night's programming with videos by trans and nb artists - if that is you, please reach out 💕

zøë wolfe

63,594 次观看 • 1 年前