Video Generation

Today's Stories

1. 1

   Microsoft Research's Mirage gives AI video a spatial memory so scenes stay consistent as the camera moves

   A team from Microsoft Research and several partner universities released Mirage on June 14, a new video generation model (software that creates video from scratch using AI) that solves a persistent problem: AI-made videos often 'forget' what a room or landscape looked like the moment the virtual camera moves away. Mirage stores scene information in a compact format called latent space, rather than building a heavy 3D map pixel by pixel, which means it uses far less computing power and graphics memory while keeping the environment looking consistent across long camera movements. The one remaining weakness is tracking moving objects — people and cars still tend to drift or disappear between video segments.

   If you use AI tools to create virtual tours, game environments, or product demos, video scenes will soon hold together much more believably as the viewpoint shifts — without needing expensive hardware to render them.

2. 2

   xAI upgrades Grok Imagine so a single photo can become a short cinematic video clip

   xAI, the AI company founded by Elon Musk, released 'grok-imagine-video-1.5-preview' on June 4, an update to its Grok Imagine image-to-video tool. Users can now upload a still photo, type a text description of what they want to happen, and the model generates a video at up to 720p resolution (roughly the quality of standard HD streaming). Multiple generated clips can be stitched together to build longer scenes.

   If you already use Grok, you can now turn product photos, travel snapshots, or illustrations into short video clips without buying or learning separate video editing software.

3. 3

   Luma AI, the video startup, opens a robotics lab that anyone can use to train real-world robots

   Luma AI, the startup known for its Dream Machine video generation tool, announced on June 1 that it is launching an open robotics lab — meaning outside researchers, students, and companies can use Luma's software to teach robots how to perform physical tasks. This is a significant pivot: the company built its reputation making AI-generated video, and it is now applying that same technology to help robots 'understand' and navigate the real world.

   Researchers or small companies that couldn't afford their own robotics infrastructure now have access to Luma's platform, which could speed up development of robots used in warehouses, healthcare, and manufacturing.

4. 4

   Az8 Studio brings a visual 'node canvas' approach to AI video production, letting creators run multiple pipelines at once

   A Reddit user's detailed first-look at Az8 Studio (posted June 8) highlighted how the tool differs from popular video generators like Runway and Pika. Instead of a single text box where you type a prompt and wait, Az8 uses an infinite visual canvas where individual tasks — creating a character, generating background audio, animating a scene — are connected as separate 'nodes' (building blocks) that pass information to each other automatically. Users can run several different creative directions side by side and compare results without starting over.

   Content creators who feel limited by the one-prompt-at-a-time approach of current video tools now have a more flexible, visual option that treats video production more like assembling Lego blocks than filling out a form.

5. 5

   Community poll reveals Veo 3 and Kling as the go-to AI video generators for everyday users right now

   A popular thread on Reddit from June 8 asked users which AI tools they actually rely on across different categories. For video generation specifically, Google's Veo 3 and Kling (from Chinese AI company Kuaishou) emerged as the two most-used tools among the respondents. Other categories showed strong preferences for Claude (text AI), ElevenLabs (voice generation), and Lovable (website building) — painting a real-world picture of which AI products have moved from hype to daily habit.

   If you're choosing an AI video tool to try today, Veo 3 and Kling are the names most commonly recommended by working users, which is more useful than relying on lab benchmarks you can't verify yourself.

6. 6

   A Reddit user finds a solid AI video setup for under $10 a month — here's what the community recommended

   A June 10 Reddit post asked for the best budget AI video workflow for generating three to four minutes of short clips daily from text prompts. Community responses highlighted that several free tiers of tools like Kling, combined with free audio tools, can cover basic needs at almost no cost, while a paid subscription to one of the mid-tier video generators typically falls in the $8–$10 monthly range. Animated and cartoon-style output was considered much easier and cheaper to achieve than photorealistic video.

   Small creators, educators, and solo business owners now have realistic options to produce usable AI video content for roughly the price of a streaming subscription — you don't need a big budget or technical skills to get started.