Welcome to the "AI Daily" column! This is your guide to exploring the world of artificial intelligence every day. Each day, we present you with the latest content in the AI field, focusing on developers to help you understand technical trends and learn about innovative AI product applications.

Fresh AI products click to learn more:https://top.aibase.com/

1. Zhipu launched a PPT generation function similar to Manus called AI Slides, free to use without limits

Zhipu has launched a new AI Slides feature based on the GLM-Experimental model, which can quickly generate high-quality PPTs based on user-provided topics or documents. The feature has clear structure, intuitive data charts, and is completely free, users can experience it through chat.z.ai.

image.png

【AiBase Summary:】

🌟 AI Slides can quickly generate high-quality PPTs based on topics or documents.

📊 The generated PPTs have clear structures and use charts to display data for easier understanding.

💻 Users can experience the AI Slides feature for free at chat.z.ai.

More details: https://chat.z.ai

2. Keling AI released Keltu 2.1 model: Significant improvement in image generation capabilities, supporting 180 styles

Keling AI released the new image generation model 'Keltu 2.1', which has significant improvements in instruction following, portrait beauty, and cinematic quality, as well as strong text generation capabilities. The new model supports over 180 styles, providing users with a richer range of creative choices.

image.png

【AiBase Summary:】

🧪 The new model performs exceptionally well in understanding complex instructions, capable of generating high-quality images accurately.

🎨 Enhanced text generation features support 180 styles, expanding creative space.

📈 Keling AI launched the Keltu 2.1 model, significantly improving image generation capabilities, users can experience for free for 7 days.

3. NVIDIA launches DiffusionRenderer: New AI model realizes video to editable realistic 3D scenes

NVIDIA and its partners launched DiffusionRenderer, a breakthrough technology that combines video generation and editing to understand and manipulate 3D scenes. The model works through neural inverse renderers and neural forward renderers, enhancing the realism and adaptability of videos and showing excellent performance in multiple tasks.

image.png

【AiBase Summary:】

🎥 DiffusionRenderer combines generation and editing functions, bringing new possibilities for 3D scene creation.

🔄 Neural inverse renderers and neural forward renderers work together to enhance video realism and adaptability.

🔧 Practical applications include dynamic lighting, material editing, and object insertion, helping creators easily perform video creation.

More details: https://youtu.be/jvEdWKaPqkc

4. Moba AI Launches: Generate High-Fidelity, Editable Prototypes in 30 Seconds by Inputting Ideas

Moba AI introduced a new prototype generation feature, allowing users to generate high-fidelity, editable prototypes from ideas in just 30 seconds, supporting multi-round dialogue optimization and local modifications, improving product design and verification efficiency.

image.png

【AiBase Summary:】

🚀 Generate editable prototypes in 30 seconds, supporting multi-terminal adaptation and multi-round dialogue optimization.

🎨 Supports multiple image inputs, intelligently parsing sketches, wireframes, etc., to generate interfaces.

🔧 Dual-mode editing and automatic document generation, achieving design as code, covering multiple scenarios.

5. Upload 10 Photos, AI Turns You into a Fashion Photo! Higgsfield Soul ID Goes Viral Globally, Revolutionizing Your Digital Self

Soul ID is a revolutionary AI tool launched by Higgsfield AI that can generate highly personalized virtual avatars by uploading more than 10 personal photos. Its core features include the perfect integration of realism and diversity, diverse style presets, and automatic prompt optimization, providing content creators and fashion bloggers with powerful creative tools.

image.png

【AiBase Summary:】

📷 Personalized training: Users need to upload more than 10 photos to generate their own AI character.

🎨 Diverse style presets: Built-in 60+ advanced style presets, switch between multiple visual languages with one click.

💡 Automatic prompt optimization: Users input simple descriptions, and AI automatically optimizes the generation conditions, outputting high-quality images.

More details: https://higgsfield.ai/

6. Google DeepMind Open-Sources GenAI Processors: One-Click Build Real-Time AI Workflows

Google DeepMind open-sourced the GenAI Processors library, providing developers with a lightweight, efficient tool for building asynchronous, composable generative AI workflows. The library supports multimodal data processing, significantly improving the development efficiency of applications based on the Gemini API.

image.png

【AiBase Summary:】

🔥 GenAI Processors simplifies the development of complex AI workflows through modular design.

🚀 Supports asynchronous stream processing of multimodal data such as audio, video, and text, improving real-time application efficiency.

🌟 Open-source community collaboration will further expand the library's functionality, covering more scenarios and programming languages.

More details: https://github.com/google-gemini/genai-processors

7. Google Veo3 Adds Image-to-Video Function, Users Create Over 40 Million Videos in Seven Weeks

Google continues to push forward in the AI video generation field, launching an image-to-video function and strengthening content recognition mechanisms, demonstrating the strong market demand for AI creation tools.

image.png

【AiBase Summary:】

🖼️ Google added an image-to-video generation function through the Gemini application, expanding the capabilities of AI creation tools.

🎥 Users can upload photos to generate video clips and add descriptive audio, supporting download or sharing of works.

🔒 All videos generated using the Veo3 model will have visible and invisible digital watermarks to ensure content traceability.

8. Mistral AI Launches Devstral2507: Designed for Code-Centric Language Modeling

Mistral AI, in collaboration with All Hands AI, launched the Devstral2507 series of models, including the open-source Devstral Small1.1 and enterprise version Devstral Medium2507. These models focus on code reasoning, program synthesis, and structured task execution, suitable for practical applications in large software codebases. Devstral Small1.1 scored 53.6% on the SWE-Bench benchmark, while Devstral Medium2507 scored 61.6%, outperforming some commercial models.

image.png

【AiBase Summary:】

🌟 The Devstral2507 series includes the open-source Devstral Small1.1 and enterprise version Devstral Medium2507, aimed at enhancing code reasoning and automation capabilities.

🚀 Devstral Small1.1 scored 53.6% on the SWE-Bench benchmark, while Devstral Medium2507 scored 61.6%, with the latter performing better than some commercial models.

💼 Both models support integration with code agent frameworks, applicable to various scenarios from local development to enterprise-level services.

More details: https://huggingface.co/mistralai/Devstral-Small-2507