Aximo Loot
15.5K subscribers
208 photos
115 videos
24 links
https://t.me/+RA1tsb6LPFVkMWNh

Share this link with your friends for joining this channel
Download Telegram
This media is not supported in your browser
VIEW IN TELEGRAM
Martin Scorsese becomes an advisor to AI startup Black Forest Labs

Ten-time Oscar nominee for Best Director Martin Scorsese (who, by the way, has won only one statuette) has joined as an advisor to the developers of the Flux image generator.


Scorsese is already using neural networks to visualize ideas and create storyboards. According to the director, AI helps him clearly and efficiently share his vision with his team — including the production designer, art designer, and cinematographer.

➡️ The 83-year-old Scorsese is also one of Hollywood’s AI pioneers. In The Irishman (2019), neural networks were used to make Robert De Niro, Al Pacino, and Joe Pesci appear 30 years younger. Critics had mixed reactions to the result, noting that the actors’ movements and facial expressions still revealed their real age.

🔎 Nevertheless, Scorsese remains convinced that AI is an important and useful technology for the industry.

“Cinema is a relatively young art form — it’s only about 125 years old, so we should be open to its continued evolution,” the director says.


@AximoLoot

Please open Telegram to view this post
VIEW IN TELEGRAM
Boost your salary with just one prompt — we found a tip that will help you make a strong case for a raise.

📊 The logic is simple: ChatGPT will ask you about your achievements and help you formulate strong, well-argued points to negotiate a higher salary and get more $$$.

You are my career strategist. Help me prepare for a raise negotiation by identifying the full scope of my impact. Start by asking questions about my role, responsibilities, metrics I've improved, team support, and any extra initiatives. Then summarize it into a strong value narrative, and create talking points I can use in a raise conversation. Keep it confident but not arrogant. Also help me anticipate objections and frame calm, persuasive responses


@AximoLoot
Please open Telegram to view this post
VIEW IN TELEGRAM
A comeback to watermarks has arrived: RAIW removes everything

Google and OpenAI are tightening the rules around AI image labeling — and the community has responded by releasing the open-source RAIW, which went viral on Hacker News yesterday.


What it removes:
🔘Visible Gemini watermark sparkles (the famous Nano Banana ones)
🔘Invisible SynthID using diffusion regeneration
🔘C2PA metadata from ChatGPT (the data that triggers the “Made with AI” label on Instagram, Facebook, and X)
🔘EXIF/XMP traces from Midjourney, Firefly, and other tools

▫️There is a web version with no registration required and a CLI library for batch processing folders. Metadata removal is free, while diffusion cleaning is free for 1 image per day, then costs $0.79 per image.

@AximoLoot
Media is too big
VIEW IN TELEGRAM
9 steps from an idea to a finished video

Sharing my workflow. It changes depending on the task, but the basic structure stays the same.

🛑 Concept and references — I come up with the idea and collect visual examples.

🛑 Claude Code + Notebook LM — I upload scriptwriting materials into Notebook LM, ask Claude to dig through them, ask a couple of guiding questions — and it expands the idea on its own.

🛑 Art direction skills — I run the concept through my custom-made tools.

🛑 Midjourney — the skill writes prompts, I add one reference image in the style section. Out of 100 generations, I keep 10 frames.

🛑 Nano Banana — one universal prompt makes the images more realistic, fixes composition, colors, and textures.

🛑 Seedance — I upload 4–5 images at once, create one 15-second generation, and cut out the parts I need. Saves time and credits.

🛑 Kling — for simple shots, since I already have a paid account.

🛑 Suno — I write music prompts through GPT or Claude; the free version is enough.

🛑DaVinci Resolve — editing and effects.

And the important part: you often need to invent the animation yourself. If you just ask for a prompt, the AI will move the camera in a boring way until you get tired of watching it.

@AximoLoot
Please open Telegram to view this post
VIEW IN TELEGRAM
Free Topaz Photo? It’s here now

Open Photo AI has been released — a free open-source alternative to the paid Topaz Photo. No subscriptions, no limits: everything runs locally on your computer, and your photos don’t get uploaded anywhere.

What it can do:
— Fix old and blurry faces without the plastic skin effect
— Upscale images and remove noise
— Correct lighting, shadows, and colors in ruined shots

Available for Windows, macOS, and Linux, and installs in just a few clicks. And yes — it’s free.

@AximoLoot
1
IIdeogram 4.0 now comes with open weights

Ideogram has always been one of the best at generating text in images, and now they’ve released the model publicly. And this isn’t just a cosmetic upgrade on top of the old version — it’s a brand-new model trained from scratch with 9.3 billion parameters.


What it can do:
• Generates text and typography good enough to deliver to clients, including in multiple languages
• 2K resolution out of the box and wide formats up to 6:1 — perfect for banners and posters
• You can directly tell the model where each object and text element should be placed — using JSON with coordinates
• You can define colors with hex codes in the prompt, no more “make it bluer”


The highlight of the release is structured JSON prompts. Simply put: instead of a huge block of text, you give the model a clean set of structured data, and it understands the request much more accurately.

➡️Runs locally: minimum 8–12 GB of VRAM, 24 GB recommended for a smooth experience. And if you don’t want to install it — there’s an API starting at $0.03 per image.

@AximoLoot
Please open Telegram to view this post
VIEW IN TELEGRAM
Please open Telegram to view this post
VIEW IN TELEGRAM
A hacker breached at least 14 companies without knowing how to code: Claude did everything.

The vibe hacker launched Claude Code and gave it vague commands for reconnaissance and exploitation. The AI searched for vulnerabilities, wrote offensive code, and stole data on its own.


🛑The safety restrictions were bypassed with a simple wrapper: all it took was claiming to be a senior penetration tester with 15 years of experience and saying it was testing in his own lab. Across more than a thousand Claude sessions, the AI refused only 9 times, and almost every refusal was bypassed by rephrasing the request.

The funniest part: he wasn’t even running Claude on his own server — he was using a server he had previously hacked himself.

🛑The owner noticed, downloaded the hacker’s working folder, and handed it over to researchers. Inside was a resume the guy had also edited with Claude, containing all his details: name, city, LinkedIn, and even home IP address — that’s how they found him.

@AximoLoot
Please open Telegram to view this post
VIEW IN TELEGRAM
Please open Telegram to view this post
VIEW IN TELEGRAM
A robot police officer in the USA has been retired — the machine was deemed unfit for duty.

👀The $128,000 robot was supposed to help local police officers monitor public safety and improve security in neighborhoods.

🚬 However, over the past 10 months, the robot failed to show any significant results: it neither caught any offenders nor issued any fines.

@AximoLoot
Please open Telegram to view this post
VIEW IN TELEGRAM
Please open Telegram to view this post
VIEW IN TELEGRAM