Google brings Gemini into video creation

Google wants video editing to feel conversational

Google says Gemini Omni can take images, audio, video, and text, then generate or edit a video from them.

The first model, Gemini Omni Flash, is rolling out to Gemini app, Google Flow, and YouTube Shorts. It starts with video, supports multi-turn edits, and includes SynthID watermarking for videos made with the model.

The most important part is the editing flow. A prompt box is less interesting than a system that remembers the scene, keeps characters consistent, and lets people revise without starting over.

This is Google putting AI video closer to everyday creation, especially through Shorts, where casual users already have a camera roll full of clips and background noise.

The avatar feature will draw attention because it lets users create videos with their own voice and likeness. Google is moving carefully on speech editing, which makes sense given how quickly trust can break around synthetic video.

The real test is whether people use Omni to finish videos, or just make impressive demos.

Figma wants AI inside the design room

Figma’s new AI agent can sit inside the same collaborative canvas where teams already argue over buttons, flows, and edge cases.

The assistant lets users generate designs, edit existing work, and run multiple agents at once through text prompts. It follows Figma’s partnerships with Anthropic and OpenAI, and arrives as the company reports $333.4 million in first-quarter revenue, up 46% from a year earlier.

This feels like the right place for Figma to push. Designers do not need another blank prompt box as much as they need help inside the messy file where work already happens.

The real test is control. If the agent understands layout, components, and product context, it could reduce grunt work without making every screen feel machine-made.

Competition from Canva, Adobe, Flora, Krea, and Dessn gives Figma a clear reason to move fast. The cursor, the side panel, and the canvas are becoming the new AI battleground.

The question is whether teams will trust an agent enough to let it touch the source file.

Stability AI wants full songs, not clips

Stability AI now says its top audio model can generate music that runs 6 minutes and 20 seconds.

The company released Stable Audio 3.0, a four-model family for sound effects, short music, and full compositions. Three models come with open weights, while the 2.7B-parameter large model sits behind paid API and self-hosting access.

That split matters. Stability is giving developers enough to experiment, while keeping the most commercially useful model under tighter control.

The timing is also practical, because music AI is moving from demo clips toward products that labels, publishers, and working musicians might actually have to judge.

The licensing claim is the key detail here, especially while Suno and Udio are still fighting lawsuits. A studio screen, a pair of headphones, and a six-minute generated track now feel less like a toy test.

The open question is whether musicians see this as useful software or another licensing fight waiting nearby.

Prohuman team

Free Guides

Explore our free guides and products to get into AI and master it.

All of them are free to access and would stay free for you.

Feeling generous?

You know someone who loves breakthroughs as much as you do.

Share The Prohuman it’s how smart people stay one update ahead.