This AI tool Made Photoshop USELESS + 18 Crazy Updates

This week, over 130 AI updates dropped, but only 19 of them are true game changers. A secret image editor that fixes only what you ask. Reals that speak Spanish or Hindi in your own voice, robots that follow English, and memory that makes your AI personal. Welcome to the AI updates of week series. Let's dive right in. Update

number one, Nano Banana, the Photoshop slayer nobody saw coming. So, there's this wild new AI called Nano Banana that just popped up on LM Marina, and it's absolutely crushing every other image editor out there. Here's the deal. Most AI editors go haywire when you try to make a simple change. Nano Banana, it nails it every time. You ask, "Swap the book cover with this image," and bam, only the cover swaps, while the hand and shadows stay exactly as they were. People are calling it surgical precision. One dude tested it by saying, "Add a third bag of dog food to the cart, same as the other two. " And yep, first try, perfect. Here's the kicker. Nobody knows who actually made this thing. No company owns it. No official announcements, no ads. It just showed up out of nowhere. Then Google's Logan Kilpatrick posted just one banana emoji. And boom, the whole AI community is like, "Yep, this has to be Google's secret project. " And it's not just hype. The numbers are insane. Want in? Head to lmarina. ai, AI. Hit battle mode, turn on image generation, drop your photo, and ask for a local edit. Keep at it until you get nano banana. You'll know by the quality. And the best part, simple prompts work best. Replace only the book cover. Keep the hand and shadows. Photoshop spent 35 years to get where it is. Nano Banana just made it look irrelevant. Update number two, Google

Pixel 10 AI that actually gets you. Your phone just got psychic. Google dropped the Pixel 10 this week. And trust me, it's not just another upgrade. This is when AI stops being a gimmick and actually starts being useful in your life. Here's the issue with phones in 2025. We've got too many apps. Planning a simple dinner? You're bouncing between messages, Gmail, maps, reservation apps, and other things. You spend more time searching for stuff than actually getting things done. Enter the Tensor G5 chip. This one's made by TSMC, not Samsung, and it changes everything. Magic Q connects the dots across all your apps, Gmail, calendar, screenshots, messages, and just knows what you need. It pulls up the relevant info and suggests actions before you even realize you need them. Tech reviewer Max Vinebach called it one of the coolest things he's seen. Android Police's editor said, "Magicu could change the way I think about AI. It was an annoyance before, but now it changes everything. I love it. " But here's where it gets wild. Voice translate breaks down language barriers during phone calls. With Tensor G5, the AI translates your conversation in real time, and you still hear the other person's voice. It's like Star Trek, but real. Tom's Guide tested it out. His voice was in English on my end, but my replies were in German on his end, and it was still my voice, not some weird assistant voice. Seriously, this is next level stuff. And the camera, it's got a coach now. Camera Coach uses Gemini models to give you tips before you even snap the shot. It'll suggest things like better lighting, better angles, or even asking your subject to look at the camera. It's like having a professional photographer in your pocket. The numbers speak for themselves. 34% faster CPU performance with huge gains in both single and multi-threaded tasks. AI tasks, a 60% performance boost. The Tensor G5 chip powers over 20 ondevice AI features right out the gate. And it's not just about phones. Google's launching Gemini for Home in October, replacing Google Assistant on speakers and displays. You can now say, "Turn off the lights everywhere except my bedroom. " And it'll actually get what you mean. No more vague commands. The Pixel 10 is available for pre-order now, and it's hitting shelves August 28th, starting at $799. It's got features that were straight up sci-fi just last year, update number three. We're at the dawn

of the memory revolution. Perplexity is rolling out memory, but a real feature currently in beta testing with plans for a full release in 2025. This memory feature learns your projects, preferences, and everyday tasks. When you ask about your ongoing project, Perplexity will recall the details from weeks ago. This shift toward personalized AI was recently confirmed by OpenAI's CEO Sam Alman in a CNBC interview. He said, "People want memory. They want product features that allow us to truly understand them. " Chat GPT's memory already got an upgrade. It now references past conversations. Ultimate called this a surprisingly great feature, noting that it points toward AI systems that not only understand you over time but become indispensable personalized assistants. GPT6, which is already in development, will focus heavily on memory. Here's a quick exercise. Create a short about me document. Include who you are, what you do, your tone, words you like or dislike, and your preferred formats. Let your AI assistants learn it. In just a week, you'll notice the difference in how much faster they get you. Update

number four, Meta Reels auto translate your voice. Any language creators, this one's for you. Meta has cracked the code to the creator language problem. They've launched an AI that clones your voice and translates your reels into any language. — Creative tools that feel almost illegal to know. Mr. Beast shared a major issue with Mark Zuckerberg. This is my most viewed video. It's just me running with progressively more and more money. And there's a reason why I don't say any words. I just go until I literally can't move anymore. It's pretty funny. Uh, and I'd love to be able to talk in this, but the problem is if I say a word, boom, like the viewership's going to get cut by more than half because, you know, then it's like only going to recommend it to people who speak English. — Yeah. No, that's super interesting. — Creators either stay silent or risk losing a global audience. Meta's new AI can autorate your reels, preserving your voice and syncing your lips. Record once in English, and the AI outputs your actual voice in Spanish. The lip movements match the translated words. The secret behind this is Meta's seamless M4T model. Early tests show 20% higher engagement for translated content. Instagram head Adam Mosery confirmed the global rollout of this feature. It's available to Facebook creators with 1,000 plus followers and all public Instagram accounts. Currently, it works for English to Spanish translations with more languages on the horizon. — As of this week, all public accounts on Instagram are going to have access to our AI translation feature. Here's how it works. Before publishing, you select translate your voice with Meta AI. You can preview the translation, approve it, and then publish. Your viewers will automatically see the content in their preferred language. The best part, processing takes minutes, not weeks. Like professional dubbing, this AI is part of Meta's No Language Left Behind project, which supports over 200 languages. Mr. Beast's problem, Meta just turned it into a solution. and global reach is just a tap away. Update

number five, sound just changed forever. 11 Labs drops video to music. 11 Labs just blew the doors of video scoring with the new video to music feature. And it's not just about matching sound to visuals. It's reading the soul of your content. Before this, scoring videos was a hassle. You'd have to search through stock libraries, pay for expensive licenses, hire composers, or settle for tracks that never really felt right. But now, you can upload any clip, and 11 Labs does the magic. A happy beach scene gets a bouncy ukulele. The AI analyzes the motion, color palette, and emotional tone of your video, then composes original music in seconds. And here's where it gets wild. You can stack everything in one place. Music, sound effects, voiceovers, all in 11 Lab Studio. Their API handles batch processing. Big players are already jumping on board. 20 major clients including film studios and gaming companies are testing it out. Ved has integrated it into their platform and AM creative agency added it to their sonic hub. One creator called it like having Han Zimmer in your pocket. This feature runs on the same 11 music engine that partnered with Cobalt and Merlin, meaning the AI is trained on licensed music, not stolen content. It delivers studiograde 44. 1 kHz output, supports any video format, and everything is commercially cleared. Want to give it a try? Head to 11labs. io/studio/videotomusic. Upload a clip and watch the magic happen. The free demo gives you 10-second tracks, while studio subscribers get full 4-minute compositions. Hold up. What's coming next is about to break the internet.

Most people won't even know this for weeks, but you're seeing it first. Trust me, the end of this video is going to get insane. Loving what you're hearing? Hit subscribe and join our WhatsApp group for exclusive AI updates and hacks. Link below. Back into the madness. Update number six, Quen image edit. Image editing just got a massive upgrade and the entire workflow is about to change. Edit first, animate later. Here's the issue with most editors. They're either locked behind pay walls or can't handle precise, detailed edits. Want to rotate a subject? Change just a t-shirt. Keep everything else exactly the same. Good luck. Quen image edit solved this. It's completely free, handles surgical edits, and is available right now. Upload any photo and ask for hyperspecific changes. Rotate the subject to front view. Replace the t-shirt with a black one that says Quen. This tool runs on a 20 billion parameter model with dual path processing. One stream understands the semantics, what you want changed. The other preserves the appearance, what stays untouched. The result, pixel perfect control over exactly what gets edited and what stays the same. Now, here's where it gets powerful for creators. Take your perfectly edited first frame and head over to Higsfield. Drop it in, draw motion paths, and hit generate. You've just created professional ad concepts in minutes instead of hours. Both platforms are live and ready to go. We're witnessing the birth of precision AI editing. It's no longer about make it different. It's about change exactly this. Touch nothing else. The future is surgical and it's free. Update number

seven, Boston Dynamics Atlas plus figure robots. Let's talk about robots. Atlas is that robot you've seen doing crazy stuff like parkour or lifting heavy things. 12 years ago, it couldn't even walk upstairs without falling. This week, Atlas is holding conversations and even reorganizing factory floors. Here's the big leap. Boston Dynamics and Toyota cracked the code. Atlas now runs on large behavior models. basically a huge neural network that can control 28 different movements at the same time. In a demo, Atlas was doing tasks like a pro, assembling stuff in a spot workshop, all from simple English commands. And when researchers purposely interfered by sliding boxes or closing lids, Atlas adapted in real time. Scott Quindesma from Boston Dynamics calls it training a single neural network to perform many long horizon manipulation tasks. Translation: It's like teaching Atlas a bunch of new tricks at once. Here's where it gets even cooler. Before, robots like Atlas had to separate walking from doing stuff with their hands. But now, Atlas treats its hands and feet the same way. One brain controlling everything. So, it can walk, crouch, manipulate parts, regrasp things, and still keep its balance. And then there's Figure 02. While Atlas figured out how to think, Figure move using reinforcement learning, basically AI learning by trial and error across thousands of virtual robots. They made the robot walk just like us. Heel strikes, tow offs, synchronized arm swings. Figure went from walking at 17% of human speed to 2. 68 mph in just one year. And these robots are already earning revenue at BMW's South Carolina plant. What's next? Atlas will be working at Hyundai factory starting in 2026. Tesla's Optimus, they're aiming to deliver 10,000 units this year. Update number eight, Google

Docs can read your document to you. Google just solved one of writing's oldest problems, catching your own mistakes. Now, Google Docs can read your documents aloud through Gemini AI. Just go to tools, audio, listen to this tab, and boom, up pops a floating player with natural sounding voices. And the best part, the voices sound genuinely human. This is a game changer for proofreading. Your ears catch things your eyes might miss. It's perfect for learning on your commute, making documents accessible, or simply multitasking. This feature powered by Gemini AI launched on August 18th and is currently available in English only. But this isn't just about convenience. It's making content consumption far more accessible. Documents now they're practically audiobooks. Sometimes the best features are the simplest ones. Update number

nine, Meta smart glasses rumor, and it's launching next month. Meta is about to change the game. Meet Hypernova. Not your typical smart glasses. These aren't just Ray-B bands with a camera. They've got a small HUD display in the right lens showing nav prompts, messages, and notifications. But the real gamecher, a neural wristband that reads electrical signals from your wrist movements. Pinch to select, rotate your hand to scroll, no touching required. Originally priced at $1,000, Meta just slashed the price to $800. Basically iPhone money for the next era of computing. At 70 g, they're a bit heavier than your current Ray-B bands, but still lightweight enough for everyday wear. Think heads-up display, not full-blown AR. The glasses will launch September 17th at Meta Connect. They're partnered with Esor Luxortica, likely branded by Oakley, since Rayban reportedly rejected the bulkier design. If Meta can nail comfort and battery life, this could be the tipping point for smart glasses to go mainstream. So, here's the big question. Will these $800 glasses with neural wristbands really replace your phone? It's a bold move or just expensive enough to fail? Time will tell. Update number 10. Google Search AI

mode gets more agentic. Google Search just leveled up. AI mode now thinks like your personal assistant. It can handle complex planning with multiple constraints. For example, find sushi for eight people near the stadium between 6:00 and 700 p. m. will give you created results with real-time availability, ratings, and direct booking links powered by Project Mariner's web browsing capabilities and Google's knowledge graph. AI mode doesn't just give you search results, it completes tasks for you. Restaurant reservations are just the start. Soon, it will handle event tickets and local services as well. Currently available to Google AI Ultra subscribers in the US via Search Labs. But here's the big news. AI mode has expanded to 180 plus countries in English. Google's shifting from here's information to I'll handle this for you. This is the dawn of do it for me search. We're watching Google evolve from an answer engine to an action engine, one task, one constraint at a time. So, you've seen the big updates this week. Now, let's rapidly go through the

fastest and wildest new AI advancements that just launched. Ready? Let's go. Runway Game Worlds lets you create interactive stories just by describing them and it generates scenes with images as you play through the adventure. ChatGpt launched a budget plan in India for under 399 rupees a month. Same AI but with 10 times higher limits than the free version. Adobe's new PDF spaces means you can finally chat with all your documents at once. Upload research papers and contracts, then ask questions across everything with proper citations. GPT5 just crushed the Pokemon Red speedrun using half the steps of older models 6,000 moves versus 18,000. That better planning translates to real workflow improvements, too. Deepseek dropped V3. 1, a massive open-source model with a reasoning toggle built right in. You can try it hosted now with local versions rolling out soon. Grammarly's new grade predictor analyzes your essay against the actual rubric and even learns your professor's grading style. Then it tells you what grade you'll probably get before you submit. Google Opal turns your app ideas into actual working prototypes. Describe what you want and it builds the workflow then gives you a sharable link. No coding required. Chat GPT users are reporting random typos in responses lately. This probably means they're testing new model routting behind the scenes. Just double check anything important. Two quick coding updates. GitHub Copilot now reads agents. mmd files to learn your project's specific rules. Excel added a co-pilot function where you can write natural language directly in cells. Which update caught your attention? Let me know in

the comments. Join our WhatsApp channel and don't forget to subscribe for next week's fire updates.