AIMarketCap | AI Trends & Prompts

Claude now works in Microsoft Word Anthropic has launched Claude for Word, built directly into your documents. What it does:

— drafts, edits, and rewrites text in Word — instantly summarizes long documents — creates content from simple prompts — works within the existing Word workflow — no need to copy and paste between apps

Chibi toy 🧸 Generate a toy based on your photo using this prompt:

3D collectible chibi-style figure attached image, ultra-detailed, stylized proportions (large head, small body), expressive face, cinematic lighting, soft shadows, Pixar-quality realism, glossy vinyl toy texture, standing pose, high detail clothing, character-accurate outfit, professional product photography, rendered in Unreal Engine 5, on a minimal studio background, toy display aesthetic, 8K ultra realistic airbrushing, or blur. All shading and highlights should be made from clean, separate color blocks. Outline all shapes with thick, bold black lines. Use a simplified color palette (such as teal, coral, mustard, and white) for a vintage feel. Add small sparkle accents or motion lines for charm, but keep the icon the clear focus. Set the illustration on a plain white background. The final result must be crisp, clean, and vector friendly with sharp edges and no texture or raster effects. Square aspect ratio.

Google integrated NotebookLM directly into Gemini 😮 You can now organize multiple projects and even add previous chats and relevant files as sources, helping you stay focused on the task at hand. This will allow users to work directly inside the Gemini interface and use Gemini chats as sources for NotebookLM. It will be available to Google AI Ultra, Pro, and Plus subscribers. Access for mobile users and free-tier users will expand in the coming weeks. More details

Retro illustration Create a retro illustration of your object using this prompt:

Create an illustration of a [OBJECT] in a retro cartoon style. Use only flat, solid colors with no gradients, smudging, airbrushing, or blur. All shading and highlights should be made from clean, separate color blocks. Outline all shapes with thick, bold black lines. Use a simplified color palette (such as teal, coral, mustard, and white) for a vintage feel. Add small sparkle accents or motion lines for charm, but keep the icon the clear focus. Set the illustration on a plain white background. The final result must be crisp, clean, and vector friendly with sharp edges and no texture or raster effects. Square aspect ratio.

Speechly — a new speech-to-text tool Speechly is more than just a transcription service — it’s a tool that adapts to the user’s speaking style, tone, and delivery. What it can do:

• optimized for macOS • adapts to your personal speaking style • makes transcription more natural and accurate • works in real time • designed for creators, developers, and everyday use

Object cutaway in Nano Banana Enter the object in the prompt below and send it to Nano Banana:

Do this for [Object] CUT PLANE: Sagittal (left-right split). Left half: intact exterior. Right half: surgical cross-section. MATERIAL_OPACITY_MAP: - Structural shell: 100% opaque exterior, revealed cross-section shows internal geometry. - Soft/organic components: 40% transparent with subsurface scattering. - Mechanical/electrical internals: Fully exposed, color-coded by function, make sure they are accurate - Fluid/gas chambers: Tinted translucent — AI infers appropriate color by function. INFER: All internal components, fastening logic, tolerances, they should be accurate STYLE: Museum specimen photography. Clean white sweep. Macro-level shadow detail. ANNOTATION: No text labels. Visual callout lines only (fine white rules).

The first OpenAI GPT-Image-2 images have leaked online 🤯 New models spotted in LMArena are already going viral:

• maskingtape-alpha • gaffertape-alpha • packingtape-alpha

Based on the first outputs, especially UI-style generations like YouTube interfaces, the model appears to have strong world knowledge and very solid text rendering. At this stage, it even looks stronger than Nano Banana Pro. Definitely one to watch.

Grok introduced “Quality” mode in Grok Imagine The new mode is powered by Grok’s most advanced image generation model. Quality mode delivers better detail, sharper text rendering, and a higher level of creative control. Details:

• a major leap in realism with photorealistic scenes, lifelike lighting, textures, and fine detail • improved text rendering with higher accuracy across many languages • deeper world knowledge, allowing the model to interpret complex scenes, realistic physics, object relationships, specific references such as brands, places, and culture, as well as richly detailed fictional or artistic worlds more accurately

It is available on the web and on mobile devices.

Cat Photo🐈 Add your photo to Nano Banana and use the prompt below

use uploaded photo as exact identity reference (no face or hair color changes); fashion editorial portrait: woman sitting confidently in oversized pink faux-fur armchair, pastel pink background; legs crossed, calm expression; fluffy pink jacket, white trousers, white lace-up boots; surrounded symmetrically by fluffy long-haired cats (cream/gray/white); monochrome pink aesthetic, soft studio light, centered composition, shallow depth, ultra-detailed, HDR, 8k photorealism

Tongyi Lab introduced Wan 2.7-Image ✨ Tongyi Lab has released an updated version of Wan 2.7-Image, a new image generation model built with a stronger focus on precision and control. What’s new:

you can now input up to 8 hex color codes to better match brand palettes the model handles multilingual text and even formulas inside images more accurately supports up to 9 reference images to keep character appearance consistent allows targeted inpainting edits and fast transparent PNG generation

In short, Wan 2.7-Image is focused not just on visual quality, but also on giving designers and marketers more reliable control over the final result. Try it here

Flowith Canvas brings people and agents together 👥 Flowith has introduced Flowith Canvas, a new workspace where you and multiple AI agents can work together on a single shared canvas. What this means:

• collaborative work between people and agents on one canvas • Claude Code and Codex agents in the same workflow • real-time coordination between multiple agents • a visual workspace instead of isolated threads

Claude Code can now use your computer 🧑‍💻 Claude Code can now control your computer directly from the command line. This is about real interaction, not just code suggestions. What it can do:

• automatically open and control applications • view user interfaces and test builds • handle write → run → debug → verify in one flow • work with local apps, Electron, SwiftUI, and graphical interfaces • supported through MCP inside Claude Code

Copilot launched the “Tasks” feature 📋 Microsoft is shifting AI from a prompt-based assistant to a task execution model. Now, you can assign a task, and Copilot will complete it without your direct involvement. What this enables:

• multi-step tasks are completed automatically • works across desktop, mobile devices, and SMS • does not require step-by-step prompting • designed for real asynchronous workflows

Lovart launched Move Object 🔳 Lovart has introduced a new tool called Move Object, turning complex image editing into a much more intuitive process. Now, you can simply select any object using a rectangle or lasso tool and move it to the desired spot in the frame. The system also adapts flexibly to additional edit requests while preserving the overall composition, helping you achieve a clean and visually consistent final image with minimal effort.

Google launched Gemini 3.1 Flash Live ⚡️ Google has introduced Gemini 3.1 Flash Live, a new voice model that responds faster, understands intonation better, and retains more context. What’s new:

• response latency reduced to 0.96 seconds • support for 90 languages • available in 200 countries and regions • context window expanded to 128K tokens • improved speech understanding even in noisy environments

The model is built on top of Gemini 3 Pro, but specifically optimized for voice interactions.

CapCut unveils Video Studio 🎬 CapCut has introduced CapCut Video Studio, a new canvas-based workspace for video creation powered by Dreamina Seedance 2.0. What’s inside:

— An AI agent for ideas, scripts, and story structure — Built-in storyboarding for scene planning — Image and video generation with multi-reference support — Full editing tools directly on the canvas

In essence, it’s an attempt to bring the entire video production workflow from idea to final edit into one window.

Claude can now control your computer Claude can now open your apps, navigate the browser, fill out spreadsheets, and do everything you do at your desktop. It can also:

— Open apps and interact with your desktop — Work with Slack, Calendar, and connected tools — Request permission to use apps without integrations — Start tasks on your phone and finish them on your computer — Automate repetitive work such as reports and email checks

Available in Claude Cowork and Claude Code on Pro and Max plans (macOS only).

HypeScribe transcribes anything in seconds 🎙️✨ HypeScribe has launched a new AI-powered platform that converts audio, video, or even a link into a full transcript, summary, and action list in under 30 seconds. 🚀 Key features:

— Paste a YouTube link or upload audio files — Instantly generate full transcripts — Create summaries and action lists — Chat with your content using built-in AI 🤖

ChatGPT lost 15% of its market share over 12 months Similarweb conducted a new study of web traffic for AI services. The research tracks the monthly traffic share of leading AI tools worldwide. Key findings: 1. As of February, Grok and Claude have overtaken DeepSeek, taking 3rd and 4th place respectively. 2. Claude surpassed the 3% mark for the first time in February. 3. Gemini is approaching a quarter of the total share. Traffic share 12 months ago vs now:

— ChatGPT: 75.7% → 59% — Gemini: 5.7% → 24% — Grok: 3.4% → 3.4% — Claude: 1.7% → 3.3% — DeepSeek: 8.5% → 3.2%

Gemini has grown almost 4× and is steadily closing the gap with ChatGPT.

▎Create vector images in 2 clicks: Lovart introduces Vectorize 🎨✨ Now you can easily convert your PNG/JPG images into full-fledged SVG vectors. Auto-tracing and outlines are also available. 🖋️ The results are ready for further work in Illustrator or Figma. 🛠️ Try it here 👇