AI just learned to build apps, create 3D worlds, and steal your face… all in a matter of seconds. OpenAI, Google, and Microsoft just dropped game-changing updates, but a new challenger from China might beat them all. We’ll show you Microsoft Edge’s new AI “Copilot Mode,” Luma’s insane video modification tool, and the latest on GPT-5. You won’t believe what’s possible now.
Here are the TOP AI breakthroughs mentioned in the video, with additional details and links to relevant articles and website.
1. Ideogram Face Swap
Ideogram has introduced a “Character” feature that allows users to create consistent characters and perform face swaps. With just a single selfie, you can place a face into various photos or generated artwork in seconds. This tool is designed for creating memes, posters, and other creative content without needing any training or a green screen.
2. ChatGPT – 5 and Study Mode and their Open source model
3. Edge Copilot Mode
Microsoft has integrated a “Copilot Mode” into its Edge browser, turning it into an AI-powered personal assistant. It can perform tasks like clicking links, booking tickets, comparing prices, and summarizing web pages directly within the browser, without the need for extensions.
- Official Announcement: https://blogs.windows.com/msedgedev/2025/07/28/introducing-copilot-mode-in-edge-a-new-way-to-browse-the-web
4. NotebookLM Video Overviews
.Google’s NotebookLM can now generate “Video Overviews” from your documents When you upload a file, it can create a narrated video with slides and Q&A support, transforming static PDFs and notes into interactive presentations.
5. GLM-4.5
Chinese AI company Zhipu AI has released GLM-4.5, a powerful new model positioned as a competitor to GPT-4. It is open-weight, self-hostable, and available for free. Benchmarks show it excels at building decks, writing code, and in some cases, outperforming models like Claude in development tests.
- https://huggingface.co/zai-org/GLM-4.5
- https://medium.com/data-science-in-your-pocket/glm-4-5-the-best-open-source-ai-model-beats-kimi-k2-qwen3-b56a5df2ec34
6. KIMI – k2
7. Google Opal App Builder
Google Labs has launched Opal, a no-code AI app builder that makes development as easy as dragging and dropping. Users can build mini-apps by chaining together Gemini prompts and functional blocks to perform tasks like scraping data, summarizing results, or sending emails, all without writing any code.
- Official Announcement: Introducing Opal: describe, create, and share your AI mini-apps – Google Developers Blog
8. Luma Dream Machine Modify
Luma AI’s Dream Machine now features a “Modify” slider that lets users rewrite their videos with text prompts. You can upload a clip, describe the desired changes, and use the slider to control the intensity of the transformation, making video editing more intuitive and accessible.
- Website: https://lumalabs.ai/
- Article: Luma’s Dream Machine, a Sora Competitor, Is Now Available to Everyone
9. Runway Aleph
10. Tencent Hunyuan 3D World
Tencent has unveiled an AI that can generate fully explorable 3D worlds from text or image prompts directly in your browser.It creates a 360-degree point of view with WASD navigation and allows for instant screenshots, functioning like a mini-metaverse generator.
- Website: Tencent Hunyuan 3D
- Article: Tencent’s AI model generates virtual worlds from a single text prompt
11. Higgsfield Halo
Higgsfield is offering unlimited AI video generation for free for a limited time through its Halo feature. This tool creates vivid, stylized 2-to-4-second video clips from simple text prompts without filters or limits.
- Website: Higgsfield
12. Fastwan
New Fastwan ai model – alibabab wan – 2.1 small model – direct use— https://fastwan.fastvideo.org/
13. Showrunner – Netflix Alternative
Showrunner website. but to create anything you have to use discord.
There are some more News that i couldn’t cover in the video, here are some of them.
- kaggle game arena v.1— https://www.kaggle.com/benchmarks/kaggle/chess-text/tournament — প্রথম AI দাবার টুর্নামেন্ট। যেখানে সব ai মডেল গুলো পার্টীসেমেট করেছে।
- Google release’s Genie 3 this is super impressive world model –https://deepmind.google/discover/blog/genie-3-a-new-frontier-for-world-models/ —এখানে text prompt , অথবা ছবি আপলোড করে দিলেইন্টারেক্টিভ 3D পরিবেশ তৈরি করতে পারে
- Skywork Unipic—https://github.com/SkyworkAI/UniPic—this is the same architecture as the legendary Gpt4o image generator
- Claude Opus 4.1-https://www.anthropic.com/news/claude-opus-4-1—claude 4.1 known to do coding benchmark is this verified
- New AI Image Editor Nano Banana is crushing the benchmark. although you can’t directly use the model yet, but there is a solution. visit lmarena and use battle mode. and activate image. from there you can upload your image and tell it what to do. the lmarena website will randomly pick 2 models and show it to you. if you are lucky, you’ll get some image output in the codename Nano Banana. of cource the name of the model will announced soon, but this is also somehting you need to know. it is better than chatgpt and gemini’s AI editing models.
Discover more from PeeAI.com - Learn AI in Bangla
Subscribe to get the latest posts sent to your email.