Adobe's MotionStream lets creators control AI video as it generates
Adobe has released a preview of MotionStream, an experimental tool that addresses a fundamental problem with AI video: the difficulty of controlling motion through text alone. The tool shifts from generating a complete video before showing it to the user, to letting creators interact with footage in real time as it's being made.
Users start with a text prompt, then click and drag objects to reposition them, adjust camera angles with sliders, and mark elements to keep static. The underlying model generates natural movement-an elephant's legs move correctly, its ears flap-without requiring manual rigging or keyframing that typically takes hours or days.
Speed and control solve two key problems
Current AI video tools force a choice between slow generation and poor control. You write a prompt, wait for the entire video to render, discover the movement looks unnatural, then start over. MotionStream removes that cycle.
The technical shift matters here. Most AI video models generate each frame by looking at every other frame-past and future-which improves quality but slows output. Adobe Research developed an "autoregressive" method where future frames depend only on what's already been created. While users watch the first section generate, the tool is already working on the next piece, enabling near-real-time feedback.
Eli Shechtman, Senior Principal Scientist on the project, said the same approach could change photo and still image editing. "Once video becomes interactive, your canvas could be a video that's always running. When you interact with it, you see a smooth video changing toward the edit you've specified."
No timeline for commercial release
MotionStream remains a research project with no announcement on whether or when it might appear in Adobe Firefly or Premiere Pro.
The release comes as questions mount about AI video's commercial viability. OpenAI shut down Sora earlier this year, and high-profile AI ads-a Coca-Cola spot and a withdrawn McDonald's Christmas campaign-demonstrated how easily generated video can look wrong to audiences. MotionStream doesn't solve whether people actually want AI video, but it does address why they struggle to use it when they try.
For creatives working with generative video or AI design tools, the shift from batch processing to real-time interaction could meaningfully change workflow speed and creative control.
Your membership also unlocks:
