Aishwary Patil on LinkedIn: #ai #deepfake #futureofcommunication #getcreative #imaginethepossibilities… (2024)

We've been heading towards this for awhile. What differentiates this from other lip sync is that it is performing responsive animation based upon content and cues - eyes blinking, head turning, and so forth. I won't get into the ethics of this - I think they should be obvious - but this is realistically what the future holds for avatars. Also note that this can be combined with TTS voice generation, so that the avatar can even sound like the targeted individual. There are still some visual clues (and I think that a mandated avatar bug should be placed in any such avatar display to indicate that this is NOT a live person), but we're reaching a point where avatar technology is becoming a thing.

25

2 Comments

Will be getting more advanced in the future. Should be interesting if there is a complementary tech to detect whether the asset is genuine or artificially generated

7

I would like to see a side-by-side with Nvidia’s existing answer to this challenge. That’s normally how these releases work.Also, the movements are likely trained on a vast corpus of data. That is good for those innocent uses, but curtails (for now) many nefarious purposes. If you want to fake a video of some “important”, you need to train and heavily bias it for that individual’s facial patterns.Right now, it would likely treat us all the same and wouldn’t fool someone who really knows the person or live up to scrutiny of the masses. That is probably a good thing at this point. I’m okay with that.It reminds me of the stories of how someone with mulitple personality disorder seems to completely transform when transitioning between personas to the point where they are almost unrecognizable. As extreme, think of a person before and after suffering a stroke. You would need to retrain your model or at least do quite a bit of transfer learning. It wouldn’t be insignificant.Now, imagine if Microsoft did for capturing individual facial patterns here what they for voice cloning in VALL-E (requires just 3 sec of random dialog). Image someone could capture just 5 seconds of video of you in a Zoom call to completely clone you.

1

1 Comment

THIS TECHNOLOGY SHOULD BE PUT ON IMMEDIATE HOLD. Talk about putting words in someone's mouth ... Just think about all the confusion this technology will cause. Theft of identity for starters. Questionability of the digitally transmitted spoken word and image. Public and private turning away from artificial intelligence and information technology. ... Realistic Heads of State avatars can be made to say anything, i.e., insult by anyone of actual Heads of State on a worldwide platform ... #informationtechnology #ai #warning #aiapplications #informationsecurity #informationmanagement #informationsecurityawareness #informationsecuritymanagement #security

6

1 Comment

AI video generation is moving fast

1

2 Comments

GAIA from Microsoft can create lifelike talking human avatars from a single photo and a voice clip.You can even add emotions or actions: like 'sad' or 'surprise,'Paper: https://lnkd.in/gWmtdgaW↓Are you technical? Check outhttps://AlphaSignal.ai to get a weekly summary of the latest models, repos and papers in AI. Read by 150,000+ engineers and researchers.

1,862

94 Comments

More and more GenAI generated realistic deep fakes we can expect, less and less we can trust what we see in the virtual space to be true. AI Act requires for deep fakes a level of transparency in order to be avare that contetnt has been manipulated. Although not technically an easy job for all modalities, it will become crucial for future trust.

2

Image+ voice to video

7

Step into the extraordinary world of GAIA, Microsoft's technological marvel that breathes life into your digital presence. Imagine transforming a single photo and your voice into a talking human avatar that not only mirrors your image and tone but also captures the nuanced dance of emotions and actions. Picture adding a touch of 'sadness' or a burst of 'surprise' to your avatar, making your digital interactions rich and deeply expressive.GAIA opens the door to a new era of personalized avatars, where every conversation becomes a canvas for genuine, human-like connection. With GAIA, your voice becomes the brushstroke, your photo the canvas, and your emotions the masterpiece. Step into a future where your digital self isn't just seen and heard but felt, creating a captivating and authentic experience in the boundless realms of the digital world.

2

1 Comment

Talking photos created with #AI are nothing new but this evolution is worth sharing to conclude the year 2023 that was overloaded with #exponential #advancements in #technology taking #humanity many steps closer to #AGI and the urgent need to talk about post labour economics.#HappyNewYear#Microsoft #GAIA#AIavatars

5