After consulting with the oracle, made some changes to the training regime and model architecture to try to enhance motion learning. Did 2 day training to 8k steps on the same dataset. It definitely learned more motion. I'm curious if it would learn even more if I did a "real" training run to 50k steps or even more, or if this is about as good as it will ever get.

#ai #aiVideo #neuralVideoCodec #video #aiTraining

Try not to laugh challengeโ€ฆ
We failed. Immediately ๐Ÿ˜‚

#BABYYUMYUM #BYY #BABYYUMYUMPREMIUM #TryNotToLaugh #SillyFaces #KidHumour #AIVideo

๐™ณ๐šŠ๐šŸ๐š’๐š ๐™น ๐™ผ๐š˜๐š›๐š›๐š’๐šœ (@david_vipernz)

@LumaLabsAI์˜ Agents๋ฅผ ์ด์šฉํ•ด Uni-1๊ณผ Ray 3.14, Grok Imagine, Suno๋กœ ๋Œ€์‚ฌ๊ฐ€ ์žˆ๋Š” ์žฅ๋ฉด์„ ์•ฝ 1์‹œ๊ฐ„ ๋งŒ์— ์ œ์ž‘ํ–ˆ๋‹ค๊ณ  ์†Œ๊ฐœํ•œ๋‹ค. AI๊ฐ€ ์˜์ƒ ํ”„๋ฆฌ๋น„์ฃผ์–ผ๋ผ์ด์ œ์ด์…˜๊ณผ ์ฝ˜ํ…์ธ  ์ œ์ž‘์„ ๋น ๋ฅด๊ฒŒ ๋•๋Š” ํ˜์‹ ์  ํ™œ์šฉ ์‚ฌ๋ก€๋กœ ๋ณผ ์ˆ˜ ์žˆ๋‹ค.

https://x.com/david_vipernz/status/2051572278564184498

#aivideo #luma #grok #suno #contentcreation

๐™ณ๐šŠ๐šŸ๐š’๐š ๐™น ๐™ผ๐š˜๐š›๐š›๐š’๐šœ (@david_vipernz) on X

For me AI isn't about replacing real actors. But if you were pitching a scene to a producer would it be better to use this method? Or words on a page? Scene made in @LumaLabsAI Agents using Uni-1 and Ray 3.14 with dialogue from Grok Imagine + Suno (1hr to make) #LumaCPP

X (formerly Twitter)

Yลki (@tanoshigee)

Flow AI๋กœ ์ด๋ฏธ์ง€๋ฅผ ๋งŒ๋“ค๊ณ , Grok์˜ ์• ๋‹ˆ๋ฉ”์ด์…˜ ๊ธฐ๋Šฅ์„ ์‚ฌ์šฉํ•ด ์˜์ƒ์„ ์ œ์ž‘ํ•œ AI ์ฐฝ์ž‘ ์‚ฌ๋ก€๋ฅผ ๊ณต์œ ํ•œ๋‹ค. ์งง์€ ๋ฐ๋ชจ ํ˜•ํƒœ๋กœ ์ด๋ฏธ์ง€ ์ƒ์„ฑ๊ณผ ์• ๋‹ˆ๋ฉ”์ด์…˜ ์ƒ์„ฑ ๋„๊ตฌ์˜ ํ™œ์šฉ ๊ฐ€๋Šฅ์„ฑ์„ ๋ณด์—ฌ์ค€๋‹ค.

https://x.com/tanoshigee/status/2051415519643800043

#flowai #grok #aivideo #genai #imagegeneration

Yลki (@tanoshigee) on X

Made a vidoe today, used Flow ai for the image, animation by @grok Inspired by @0xInk_ Cool right?

X (formerly Twitter)

Artificial Analysis (@ArtificialAnlys)

Video Rebirth์˜ Bach-1.0 Preview๊ฐ€ Artificial Analysis ํ…์ŠคํŠธ-ํˆฌ-๋น„๋””์˜ค ๋ฆฌ๋”๋ณด๋“œ์—์„œ 6์œ„๋ฅผ ๊ธฐ๋กํ•˜๋ฉฐ ๊ณต๊ฐœ๋˜์—ˆ์Šต๋‹ˆ๋‹ค. ๋ฌด์Œ ํ…์ŠคํŠธ-ํˆฌ-๋น„๋””์˜ค ๋ชจ๋ธ์˜ ์ตœ์‹  ํ”„๋ฆฌ๋ทฐ๋กœ, Vidu Q3 Pro, Kling 3.0 Omni 1080p(Pro), grok-imagine-video์™€ ๋น„์Šทํ•œ ์„ฑ๋Šฅ์„ ๋ณด์ธ๋‹ค๊ณ  ์†Œ๊ฐœ๋ฉ๋‹ˆ๋‹ค.

https://x.com/ArtificialAnlys/status/2051236578455110010

#texttovideo #aivideo #videomodel #leaderboard #generativeai

Artificial Analysis (@ArtificialAnlys) on X

Bach-1.0 Preview from Video Rebirth debuts at #6 on the Artificial Analysis Text to Video Leaderboard (No Audio)! Bach-1.0 Preview is the latest Text to Video model from @video_rebirth, with similar performance to Vidu Q3 Pro, Kling 3.0 Omni 1080p (Pro), and grok-imagine-video.

X (formerly Twitter)

Mark Gadala-Maria (@markgadala)

AI ๋น„๋””์˜ค์˜ ๋œ ์ฃผ๋ชฉ๋ฐ›๋Š” ํ™œ์šฉ ์‚ฌ๋ก€๋กœ ๊ฒŒ์ž„ ์ฝ˜์…‰ํŠธ ๋””์ž์ธ์ด ์ œ์‹œ๋๋‹ค. GPT-Image 2๋กœ ์Šคํฌ๋ฆฐ์ƒท ์Šคํƒ€์ผ์˜ ์žฅ๋ฉด์„ ๋งŒ๋“ค๊ณ  Seedance 2๋กœ ์ด๋ฅผ ์• ๋‹ˆ๋ฉ”์ด์…˜ํ™”ํ•˜๋Š” ์›Œํฌํ”Œ๋กœ๊ฐ€ ์†Œ๊ฐœ๋˜๋ฉฐ, ์ƒ์„ฑํ˜• ์ด๋ฏธ์ง€์™€ ์˜์ƒ ๋„๊ตฌ์˜ ๊ฒฐํ•ฉ ํ™œ์šฉ ๊ฐ€๋Šฅ์„ฑ์„ ๋ณด์—ฌ์ค€๋‹ค.

https://x.com/markgadala/status/2051348475359449101

#gptimage2 #seedance2 #aivideo #gamedesign #genai

Mark Gadala-Maria (@markgadala) on X

One of the underrated uses of AI video right now is video game concept design. Use GPT-Image 2 for screenplay screenshots and then animate them using Seedance 2.

X (formerly Twitter)