🤖 AI Weekly Recap (Week 4)

Happy Sunday! We just had another crazy week in AI. China just dropped a free AI voice model that clones any voice in 3 seconds while Krea’s new tool lets you edit images in real-time.

And that's not all, here are the most important AI moves you need to know this week.

1. Anthropic launches Claude in Excel

Anthropic has opened access to Claude’s improved Excel integration for all Pro subscribers, moving it out of limited beta and enabling AI-powered spreadsheet analysis directly inside Microsoft Excel.

• Work with Claude directly inside Excel for analysis, editing, and data preparation

• Drag and drop multiple Excel files into a single Claude session

• Prevents overwriting existing cells to reduce accidental data loss

• Supports longer work sessions through automatic context compression

• Available as an official Excel add-in via the Microsoft Marketplace

Try it Now → https://claude.com/claude-in-excel

2. Alibaba releases Qwen3-TTS

Alibaba has released Qwen3-TTS, a free open-source text-to-speech model that can clone any voice using just 3 seconds of audio while supporting 10 languages and 49 different character voices.

• Clone voices from as little as 3 seconds of reference audio

• Includes the full Qwen3-TTS family: VoiceDesign, CustomVoice, and Base models

• Available in 5 model variants ranging from 0.6B to 1.8B parameters

• All weights, code, and the research paper are fully open-sourced

Try it Now → https://github.com/QwenLM/Qwen3-TTS

3. Z.ai Releases GLM-4.7-Flash

China’s Z.ai has released GLM-4.7-Flash, a free and efficient 30B-class model designed for local deployment, coding, and agentic workflows, while also performing strongly across creative and long-context tasks.

• Sets a new performance standard for the 30B model class

• Optimized for coding, agentic reasoning, and tool-use workflows

• Outperforms Claude Sonnet and GPT-5 on multiple coding benchmarks

• Runs locally at zero cost or via hosted access for around $3 per month

• Balances high reasoning performance with low compute and memory requirements

Try it Now → https://chat.z.ai/

4. Runway Releases Gen-4.5

Runway has released Gen-4.5, the #1 ranked text-to-video model globally, surpassing Google and OpenAI with unprecedented physics accuracy, cinematic quality, and full-HD video generation at 1080p and 30fps.

• Top-ranked on the Artificial Analysis Text-to-Video leaderboard with 1,247 Elo

• State-of-the-art motion quality, prompt adherence, and visual fidelity

• Generates cinematic, highly realistic video at 1080p and 30fps

• Delivers realistic physics including weight, momentum, collisions, and fluid dynamics

• Available across all Runway subscription plans at comparable pricing

Try it Now → https://app.runwayml.com

5. LTX Launches Audio-to-Video Generation with ElevenLabs

Lightricks has launched audio-to-video generation inside LTX in partnership with ElevenLabs, allowing creators to turn voice, music, and sound effects directly into cinematic video, starting from sound, not text.

• Generates video directly from audio, with sound acting as the control layer

• Voice cadence drives pacing, motion, and scene transitions

• Music energy influences camera movement and visual intensity

• Supports voice, dialogue, music, and sound effects as primary inputs

• Optional image input to anchor characters or scenes

Try it Now → https://ltx.studio/

6. Krea AI Unveils Real-Time Complex Image Editing

Krea AI has introduced Realtime Edit, a new beta feature that lets users paint, sketch, or describe changes and watch images transform instantly, enabling true “edit-as-you-direct” image creation.

• Instant real-time updates with AI interpreting inputs in milliseconds to seconds

• Sketch-to-image editing that turns simple drawings into high-fidelity visuals

• Text-based edits to modify objects, composition, lighting, and shadows on the fly

• Live webcam and camera feed support for real-time AI effects

Try it Now → https://www.krea.ai/realtime