Latest news with #Captions
The Verge
5 days ago
- Entertainment
- The Verge
We are not ready for better deepfakes
If you're like me, then lately you've scrolled past something on social media and thought, 'Wait, was that real?' Deepfakes are everywhere, and they're getting a lot more convincing. That brings me to my Decoder guest today: Gaurav Misra, the CEO of Captions. You may not have heard of Captions yet, but you've probably seen a video that was generated using its AI models. The company's Mirage Studio platform lets anyone generate AI versions of real people, and the results are alarmingly realistic. Captions just put out a blog post titled, 'We Build Synthetic Humans. Here's What's Keeping Us Up at Night.' It's a good overview of the state of deepfakes and where they're headed. As the CEO of a company building deepfake technology, I wanted to know what specifically keeps Gaurav up at night, which you'll hear us get into. I'm generally more optimistic about the long-term impacts of AI than a lot of people, but as you'll hear in this conversation, I'm a lot more nervous about this topic. Ultimately, I came away from this episode unsettled by the fact that the deepfakes of today are the least believable they'll ever be, we are not ready, and the companies building this tech are racing ahead anyway. If you'd like to read more on what we talked about in this episode, check out the links below: Questions or comments about this episode? Hit us up at decoder@ We really do read every email! A podcast from The Verge about big ideas and other problems. Posts from this author will be added to your daily email digest and your homepage feed. See All by Alex Heath Posts from this topic will be added to your daily email digest and your homepage feed. See All AI Posts from this topic will be added to your daily email digest and your homepage feed. See All Decoder Posts from this topic will be added to your daily email digest and your homepage feed. See All Podcasts Posts from this topic will be added to your daily email digest and your homepage feed. See All Tech
The Verge
23-07-2025
- Entertainment
- The Verge
A look at the state of deepfakes.
From ChatGPT to Gemini: how AI is rewriting the internet See all Stories Posted Jul 23, 2025 at 12:00 AM UTC A look at the state of deepfakes. If you're interested in a high-level overview of the state of deepfakes, I recommend checking out this blog post from Captions, a Capcut competitor and a software platform that enables creators to generate AI-generated videos from scratch. What struck me the most is the company's prediction that, 'Very soon, most models will allow you to generate a person (real or synthetic) in any situation – without a duration constraint (longer than 12 seconds) and featuring multiple people in one shot.' Chat, are we cooked? I'll have Captions CEO Gaurav Misra on Decoder later this week to talk about this. Follow topics and authors from this story to see more like this in your personalized homepage feed and to receive email updates. Alex Heath Posts from this author will be added to your daily email digest and your homepage feed. See All by Alex Heath Posts from this topic will be added to your daily email digest and your homepage feed. See All AI Posts from this topic will be added to your daily email digest and your homepage feed. See All Creators Posts from this topic will be added to your daily email digest and your homepage feed. See All Tech
India Today
12-06-2025
- Entertainment
- India Today
Meta AI can now edit videos, but with limited preset prompts
Meta has unveiled a new suite of AI-powered video editing tools, expanding the capabilities of its Meta AI platform. The latest features allow users to edit short videos using preset artificial intelligence prompts that can transform aspects such as outfits, settings, and overall styles, a move that puts Meta in direct competition with rivals like Google, which has been steadily introducing generative AI video features, alongside other specialist editing platforms such as Captions. advertisementThe company announced on Wednesday that these new tools will be made available on the Meta AI app, the website, and its standalone video-editing app, Edits, Meta's alternative to popular apps like CapCut. For now, the rollout is limited to users in the United revealed that the inspiration for these editing features stemmed from its Movie Gen AI models, although it has not been confirmed whether these models are being directly used to power the new editing functionality. At present, users will have access to 50 preset options, which can be applied to video clips up to 10 seconds long. According to Meta, these presets were designed following feedback from creators, aiming to offer an easy and accessible editing experience that reflects real-world creative needs. The available presets include a wide variety of effects, for example, users can apply a 'vintage comic book style', simulate a rainy day lighting effect, or change a subject's outfit to a space cadet suit. advertisement Once edited, the videos can be shared directly across Meta's major platforms, including Facebook and Instagram, allowing for immediate distribution to followers and the current version provides a limited selection of presets, Meta has confirmed that it intends to introduce additional customisation options later this year, further enhancing creative possibilities for users. The company stressed that the tools are designed to make video editing simple, creative, and accessible to everyone. 'We built this so that everyone can experiment creatively and make fun, interesting videos to share with their friends, family, and followers,' Meta stated in its blog post. 'Whether you're reimagining a favourite family memory or finding new ways to entertain your audience, our video editing [tools] can help.'These new video-editing capabilities build upon Meta AI's existing image generation tools, which are already integrated across several of Meta's platforms. The addition of video editing suggests a clear attempt to encourage content creators to remain within Meta's own ecosystem, rather than turning to third-party apps for advanced editing move reflects a wider industry trend, as major technology firms race to incorporate generative AI into their platforms to meet the growing demand for content creation tools. As short-form video continues to dominate social media engagement, Meta's expansion into AI-driven video editing positions the company to better serve creators looking for efficient, innovative, and user-friendly editing Google, Captions, and other competitors advancing their own AI-powered video capabilities, Meta's latest update signals its ambition to remain at the forefront of the rapidly evolving content creation landscape. By making professional-style video editing accessible to a broader audience, Meta aims to empower both casual users and professional creators to produce dynamic, eye-catching content without the need for complex software or advanced editing skills.
Gizmodo
04-06-2025
- Business
- Gizmodo
This AI Company Saw Google's Veo 3-Powered YouTube Slop and Said, ‘Hell Yeah'
The video generator train has left the station and there's no turning back. Video generation is all the rage right now. While last month was all about Google's Veo 3, which has already threatened to deepfake all of YouTube, make incoherent action movies, and automate game development, the fun (or horror) hasn't stopped there. Just this week, OpenAI announced that it's making its Sora video generation model available for free via Bing, which is big news for anyone that's interested in testing out a video generator but doesn't actually feel the need to pay for one yet. And it's not all about the big two, either—there are apparently other AI companies that saw what Veo 3 was putting down in the world of AI slop and said, 'Hold my beer.' One of those companies is called Captions, which just launched its own video generation model called Mirage Studio that it says can 'generate expressive videos at scale' and create 'actors that look and feel alive.' To translate: it looks like Mirage Studio is designed to be the perfect model for AI-generating social media slop, and semi-convincing AI social media slop at that. Introducing Mirage Studio. Powered by our proprietary omni-modal foundation model. Generate expressive videos at scale, with actors that actually look and feel alive. Our actors laugh, flinch, sing, rap — all of course, per your direction. Just upload an audio, describe the… — Captions (@getcaptionsapp) June 2, 2025 Mirage Studio seems, unlike Veo 3, to be almost singularly geared toward 'content creation,' which is to say videos that might appear on TikTok, Instagram, or YouTube. As Captions puts it, Mirage Studio is 'built for marketers, creative teams, and anyone serious about crafting great narrative videos.' It works by uploading an audio clip and then either describing a scene to generate or dropping in a 'reference image.' Once all of that is in there, you just select a few parameters, and boom, you generate the video. If this sounds like an easy AI-ified way to steal the likeness of content creators or actors, I'm here to report that that's exactly what it is. As proof, I took some of Gizmodo's content from my colleague Kyle Barr (sorry, Kyle), who got to check out the Switch 2 recently, and turned his work into a Twitch-like abomination delivered by this guy with a blue mohawk. It's hard to say who Mirage Studio is stealing from (aside from us), but I'd hazard a guess and say it's clearly trained on videos from YouTube, Twitch, and other streaming platforms. To make the video, I did exactly what Captions said: I took a video from Gizmodo's Instagram page, threw it into the AI slopifier, selected my 'actor,' and waited for Mirage Studio to melt it down inside its generative belly and spit it back out. After an excruciating 10 minutes of waiting for a three-second preview, I finally got a little clip of what Kyle's video would look like if delivered by a blue-haired Twitch streamer, and then I got to generating the video in full. The results are about exactly what I might expect from an AI video generator of this caliber. I turned @KyleBarr5's Switch 2 video on Gizmodo into AI slop with @getcaptionsapp. Sorry, Kyle, and everyone with eyeballs. — James Pero (@jamestpero) June 4, 2025 At a very cursory glance, it did an okay job, but the more you watch, the more it unravels. The narration is mostly fine since it basically stole Kyle's delivery and regurgitated it, but there are hazy edges around the perimeter of the video and other small visual distortions. Also—my personal favorite—a spontaneous appearance of purple ectoplasm toward the end, which I assume is some kind of hallucination of an object from another video that Mirage Studio stol—er—borrowed from. It's a little sloppy and more than a little lazy, but it's also (unfortunately) more than a little discouraging. It's pretty clear from this small demo that the AI slop train is coming, and there's little we can do to prevent it. Outside of implications on content and the deluge of mindless social media chum we may encounter as a result of video generators, there are also big questions we need to unpack about intellectual property—questions with answers that don't seem to be forthcoming anytime soon. Hollywood actors have already urged regulators en masse to set a framework for this sort of thing, but as is the case with most tech policy, it seems they're content with being reactive as opposed to preemptive. In any case, Mirage Studio is just further proof that video generators are coming (or they've already arrived), and that's probably great news if you want to churn out some BS video for clicks, but for anyone that puts time and effort into making content, it may be time to buckle up because things are about to get, well… sloppy.
The National
22-04-2025
- Business
- The National
Meta releases Edits app as company battles FTC in court
Meta released a video app called Edits on Tuesday, as the social media company continues to face off in court against the Federal Trade Commission. The debut of Edits, which seeks to increase Meta's already massive social media footprint, comes at one of the most challenging times in the company's history, with US federal regulators accusing it of engaging in anticompetitive behaviour. Last week, Meta chief executive Mark Zuckerberg took the witness box to combat the accusations. The trial is expected to last at least two months. Meta, which owns Facebook, Instagram and WhatsApp, has been touting the release of its mobile video editing tool for months. Edits will compete in the same space as Captions, CapCut and other apps used by social media creators. 'If you're passionate about making videos, Edits has the tools you need to support your creation process, all in one place,' Meta said in a news release. 'This is just the first step in making an app that helps you create your best videos, and we plan to keep evolving and improving Edits based on your feedback.' Video creation has become paramount for Meta, as it seeks to blunt the impact and rising popularity of video-based apps like TikTok. Several years ago, the company pivoted its Instagram platform to begin showcasing more videos – called Reels – instead of traditional photo carousels, which previously made up a majority of the app's content. Those efforts have been relatively successful, with Instagram proving to be a worthy competitor to TikTok, although it has yet to overtake the China-owned app in popularity. Edits claims to allow users of Meta platforms to create better content at a quicker pace, with captions, transitions and special effects able to be uploaded in various places. Meta has touted powerful editing tools within the app such as 'frame-accurate timeline with clip-level editing, auto-enhance features and effects like green screen and transitions'. The company also says that the new app provides users with 'real-time feedback on factors that can affect distribution', so that users can change videos to increase visibility. Meta's purchases of Instagram and WhatsApp are under particular scrutiny.
