Latest news with #KatelynChedraoui
CNET
4 days ago
- Entertainment
- CNET
How To Fix the Most Common AI Image Errors and Hallucinations
One of the best parts of my job is laughing at the truly terrible, deeply flawed, occasionally frightening AI images that pop out while I review different AI image generators. I spend a lot of time playing around with AI in ChatGPT, Midjourney and Adobe, and I've learned that no AI generator is perfect. But there are some common patterns, things that many AIs struggle with. Every image generator is unique, but I've spent enough time with them to recognize some patterns. There are certain things AIs are prone to mess up -- fingers on hands, overlapping elements, extremely fine details, etcetera. Some AI image programs give you tools to edit these mistakes, which is great. But when you can't, these are my tried-and-true tips for tweaking your prompts and settings to fix your images. These are the biggest problems I found and how to fix them. For more, check out the best AI image generators and our guide to effective AI image prompt writing. Human faces and expressions Katelyn Chedraoui via Canva Magic Media AI/CNET Accurate facial expressions continually challenge AI generators. Quirky eyes, teeth and eyebrows are some of the strongest indicators that an image is AI-generated. In this case, the result was extremely funny to me, if also completely unusable. The girls are sporting some Halloween-like vampire teeth, and the dude in the back is having more than a bad hair day. Stephen Shankland via Dall-E 3/CNET Even with cartoon or non-realistic characters, generators struggle to moderate emotion and expressions. This image -- created by our best pick, Dall-E 3 -- over-amplified the prompt, and the end result was too dramatic. I'm a self-identified neat freak, but I can't imagine anyone getting this upset over what looks like hundreds of dollars of cleaning supplies. Even the best programs can fail and produce wonky results. How to fix it: I recommend asking the service to cut down on the number of people it's trying to render -- cut down on the number of chances for error -- and using post-generation editing tools to select specific parts of the image that need regenerated or fixed. Picking a more mild adjective ("angry" rather than "enraged") might help guide the service down the right path. Iconic characters, trademarks and recognizable logos Katelyn Chedraoui via Canva Magic Media AI/CNET For many reasons, AI image generators can rarely produce accurate logos, known trademarks or otherwise recognizable characters. This is usually because there are significant legal concerns about a company allowing its brand materials or intellectual property to be used for AI services. It also could be because the logo or brand you want to include isn't in the generator's training data, so the generator won't understand your request. Two recent exceptions to this are new image generation features on the Google Pixel 9 and Pixel 9 XL, thanks to Gemini AI, which our mobile experts used to create images featuring accurate-looking Mickey Mouse and Pikachu renditions. The other exception is some paying X (formerly Twitter) users reporting that they can use the social media site's Grok chatbot to make realistic-looking Mickeys and human political figures. These images still aren't perfect renderings, but they're pretty close. How to fix it: You can't, but on the plus side, you're a lot more likely to avoid lawsuits for copyright infringement. If you run into this issue, I recommend rethinking your design concept to see if you can get the same point across without using a specific company name or logo. Do you really need the TikTok logo, or do you just need a phone with a vertical video playing, for example. Overlapping, complex elements Katelyn Chedraoui via Leonardo AI/CNET Occasionally, you'll stump a generator if you have too many overlapping or complex elements within an image. The best AI image programs aren't going to run into this issue very often, but even the best programs fail at some point. With this Leonardo image, the library of my dreams is ruined by the rolling ladder that disappears halfway up. Katelyn Chedraoui via Canva Magic Media AI/CNET Similarly, this photorealistic image of a kitchen looks alright from a cursory look, but a second glance or quick zoom-in shows nonsensical characters printed on the cookbook, and the book itself seems to have two spines and three sections. Images like this show how frustrating it can be to create AI images -- even when you think you've got a winner, small flaws make them unusable. How to fix it: Try simplifying your prompt or using post-generation editing tools to single out the problem and fix it. Sometimes changing the aesthetic can help as well; these issues usually pop up when you're trying to make photorealistic or stock images. If your service offers a specific area editing tool, try using that to select the questionable area and ask the generator to remove what's in it. Over-editing and hallucinations Katelyn Chedraoui via Midjourney AI/CNET The sign of a great AI image generator is its ability to produce an image with few errors or hallucinations while giving you the right tools to fix those issues when they inevitably pop up, but not every service does. In this example from Midjourney, I went through several rounds of edits on this image of a soccer team celebrating a victory, and I have no idea what happened to get to this final result -- or any clue what's going on with the blob that's presumably a player in the right-hand side of the picture. Frankly, I don't think Midjourney knew either. How to fix it: With editing AI images, sometimes less is more. Don't be afraid to scrap your current batch of images and start over. You can often preemptively fix big issues by refining your prompt so you only have to worry about fixing minor issues later. AI generators still need human work Katelyn Chedraoui via Leonardo AI/CNET I have no doubt that the companies behind these AI generators are working to eliminate these issues and level up their models. For now, imperfect images show us once again that while this tech is improving, it's not flawless. My hope is that because so many of these services face similar challenges, they will all be chasing solutions for these issues. As a final note, I'll remind you that it's a best practice to credit or acknowledge that an image is AI-generated when you share it. As AI images get better and more realistic-looking, it's more important than ever to distinguish AI-generated art from other kinds of media. For more, check out our advice for writing the best AI image prompt and our best picks for AI image generators.
CNET
15-05-2025
- Business
- CNET
Meta Reportedly Delays 'Behemoth' AI Model: What This Could Mean for Its AI Tools
Meta has reportedly pushed back the release of its Behemoth large language model for its artificial intelligence tools, delaying it until the fall. Behemoth was originally planned to release in April to coincide with Meta's first AI conference, LlamaCon, but was delayed until June before being delayed again now, according to a report by The Wall Street Journal on Thursday. Meta released Llama 4 in April. Llama -- Large Language Model Meta AI -- is Meta's family of LLMs. But Meta AI engineers are concerned the capabilities in the Behemoth LLM aren't a significant enough improvement over what's already available via Llama 4 to warrant a release next month, according to WSJ citing unnamed sources within Meta. A large language model is what sits behind the chatbot interface you interact with when you ask AI tools like ChatGPT, Gemini and Meta AI a question. LLMs understand and generate human-like text. Behemoth is "one of the smartest LLMs in the world and our most powerful yet to serve as a teacher for our new models," Meta said in April. The tech giant aims to become one of the biggest AI providers, and has already woven AI into how you interact on many of its apps across Facebook, Instagram, WhatsApp and Messenger, including helping with writing posts and captions, and editing your images. Meta also released a standalone app for Meta AI at the end of April, which includes a hub for its Ray-Ban smart glasses. CNET social media and AI reporter Katelyn Chedraoui, who reported on LlamaCon in April, said this fresh delay adds to concerns that Meta is already behind the curve on what its AI tools are capable of. "Meta's reported decision to delay the release of its new Llama models means the company could fall even further behind other big companies like OpenAI and Google," Chedraoui said. "The race to build advanced and affordable AI is extremely competitive. The concern for Meta is how far ahead competitors will advance while it's still working on refining the models it has already announced." Meta didn't immediately respond to a request for comment.
CNET
14-05-2025
- Entertainment
- CNET
I Tried TikTok's New AI Video Generator. It's All Kinds of Glitchy
AI is encroaching on a new area of social media. TikTok on Tuesday announced the new feature AI Alive, which lets creators animate photos for TikTok stories into short video clips using generative AI. The feature is rolling out in the US this week. My experience using AI Alive has been a mixed bag. The first time I tried it, it took nearly 5 minutes to generate a video, and for 2 of those minutes, it kept saying there were only 10 seconds left. The results ranged from acceptable, at best, to downright horrifying. TikTok also seemed to struggle with object permanence. These are a couple of the videos I made with the new TikTok feature. In the first, I wanted to add a dog to my picture of sunny Tybee Island in Georgia, but TikTok took the wheel and decided to make the dog fly in from nowhere, instead of using my directive in the prompt to have the dog run in on the ground. Created by Katelyn Chedraoui using TikTok AI In another video, AI replaced existing elements, unprompted. My original poster of a ram became something of a horror scene, with the now rather scary-looking beast leaping through the page -- I'd simply instructed AI Alive to animate the existing ram and have it jump off the poster. And I don't know why TikTok's feature ultimately decided to replace the original poster with weird AI gibberish. That was unprompted, and the final frame of the video kind of reminds me of an ad for a scary-clown movie. Created by Katelyn Chedraoui using TikTok AI There were lots of weird glitches throughout my testing. I think I could've fixed some of those blunders with edited prompts, but you get only five attempts per day, a disappointingly low limit. The AI video model powering the new feature was built by TikTok, with certain elements supported by open-source technology. Given that TikTok is first and foremost a social media company, not an AI developer, it makes sense that it's taking advantage of some open-source AI tech. TikTok's AI Alive videos will automatically have a label added to them denoting that they're AI-generated, and as part of TikTok's partnership with the Coalition for Content Provenance and Authenticity (C2PA), similar markers will be embedded in the metadata of the videos. While the feature itself is meh, this announcement tells us about what TikTok imagines as its future with AI. Many social media platforms have undergone an AI makeover over the past few years, none more so than Instagram and Facebook with Meta AI. TikTok has launched some AI features, including avatars for professional creators and behind-the-scenes AI tools for marketers. But this is its first general-purpose and widely available AI feature. It could mark the beginning of TikTok's AI era. For a company that's been mired in privacy concerns for years, fueling federal legislation, a Supreme Court case and multiple presidential stays to keep the app online in the US, TikTok's choice to dive deeper into AI is noteworthy. Generative AI has its own myriad of privacy concerns, and for image and video generators, people often raise questions about how the models are trained and how they process users' data. TikTok's privacy policy doesn't have a dedicated AI section, but it says TikTok can use information you give it to "improve and develop the platform and conduct product development," which could include AI models. I'm struggling to imagine an essential use case for this feature for TikTok creators, but it's probably worth playing around to see if it could work for you. If you want to try it out, here's how to get going: How to animate your TikTok photos with AI Open TikTok and navigate to your profile. Tap the blue plus sign on your profile pic. Upload a photo or take a new picture using the camera. On the left side menu, tap AI Alive. It's under the text option, above stickers. Write your prompt. Choose one of TikTok's suggestions for the best results. Finish editing your video however you want. Post your story or save it to your camera roll. This is one of the better TikTok AI videos I made. Created by Katelyn Chedraoui using TikTok AI Writing a good AI prompt is the key to success. This is true for all AI image and video generators, but in my testing, TikTok's AI seems especially sensitive. When you're writing your prompt, put the most important elements at the beginning. You can check out our full guide to writing effective AI image prompts for more tips. For more, check out these CapCut alternatives and our top picks for AI image generators.
CNET
13-05-2025
- Entertainment
- CNET
Instagram Video Editing App Update Includes New Features to Rival CapCut
The very real threat of a TikTok ban has caused Americans to search for alternatives to their favorite ByteDance-owned apps -- including the popular video editing software CapCut. So it's no surprise that the recently launched Instagram Edits is capitalizing on consumer uncertainty by adding features from one of its most popular competitors. The latest update for the video editing app added a host of new features to better compete with similar pieces of software. These features mix existing technology with Meta's algorithmic data to create a hybrid experience between social media and video editing. The beat editing feature will help users align their audio and visual elements without a hassle. Instagram Arguably, the most important addition to the app is beat editing. What is beat editing? Beat editing is a feature that CapCut has had for years, but its inclusion in the Instagram Edits app shows that Meta wants to seriously rival the ByteDance video editor. This feature automatically creates beat markers on the editing timeline that helps you line up your audio with your visual components (clips, text and overlays) when merging them together for a video. You'll now have access to a new set of preset animations and visual changes that can be applied to your videos -- many of these emulate shaky recordings, change color saturations or make similar changes to the final product. CNET's Katelyn Chedraoui included Instagram Edits on a list of the best CapCut alternatives, calling it "frankly a CapCut copycat." Instagram Edits is also testing more social features, implementing a social media approach to the video editing app. The new Inspirations tab allows you to browse Instagram Reel edits made by people the algorithm deems as being similar to you. From here, you can save trending audio you find to use in your own Reels, or bookmark certain clips you find for later use. A representative for Meta, the company that owns Instagram, did not immediately respond to a request for comment.
