Text to Video
Start from a written shot brief when there is no source asset yet. The prompt focuses on subject action, scene, camera movement, and pacing.
AI video workspace
Generate videos from text, images, and reference frames.
Plan the shot like a real video tool: choose text-to-video, image-to-video, or reference mode, add source frames, control camera movement, set motion strength, duration, and aspect ratio, then preview clips in one workspace.
Mode
Text
Duration
8s
Motion
Natural
Results
Preview studio ready
TextPush inNatural8s
Preview
Live sample playback from the Muse Video studio.
Built around the way video generators actually work.
Runway, Pika, Luma, Kling, Hailuo, and Vidu all separate source mode, reference material, and motion controls. Muse Video now follows that mental model.
Image to Video
Use a first frame or product image as the visual anchor, then add camera and motion direction without treating the task like image generation.
Reference frames
Bring multiple frames for character, product, style, or first-last-frame continuity so the model has visual context before motion is added.
Video scenarios, not image batches
Examples are organized by motion intent across product, portrait, social, cinematic, beauty, culinary, travel, and interior shots.
Choose the right generation mode first
Text-only ideas, first-frame animation, and reference-driven continuity have different inputs. The studio exposes that choice before the prompt so users do not have to guess what the model will use.
Start generatingText to Video, Image to Video, Reference frames
Source frames are first-class inputs
Image-to-video and reference workflows show upload states, frame roles, and source-frame metadata instead of pretending every job starts from text.
Start generatingFirst frame, last frame, product reference, character reference
Camera and motion are explicit controls
Real video generation needs push-in, pan, orbit, handheld, static, subtle, natural, and dynamic choices close to the prompt, not buried as vague style text.
Start generatingPush in, orbit, pan, handheld, subtle motion, dynamic motion
Preview and iterate like a video tool
The player, current brief, source frames, and generation history stay on screen so each clip can become the next shot direction.
Start generatingPreview, reuse prompt, copy final brief, download clip
How it works
The flow mirrors real video ideation: choose the source mode, add visual anchors when needed, set motion controls, then generate and refine.
Pick the source mode
Start with Text, Image, or Reference mode so the workspace knows whether this is pure text-to-video, image-to-video, or a visual continuity task.
Add frames when needed
Image and reference modes expose source-frame upload and frame roles before generation, which matches how modern video models are prompted.
Direct camera and motion
Set camera movement, motion strength, ratio, and duration next to the prompt so every generated clip has an intentional shot brief.
FAQ
The surface is now planned around video workflows while keeping a production-ready provider architecture.
Can Muse Video do both text-to-video and image-to-video?
Yes. The workspace now has separate Text, Image, and Reference modes so pure prompts and uploaded source frames are handled as different video workflows.
Can I use more than one reference frame?
Reference mode supports multiple local source frames in the UI. A production provider integration can map those fields to provider-specific reference, first-frame, or last-frame inputs.
Can it connect to a real video provider?
Yes. The workbench is structured for text-to-video, image-to-video, and reference-frame provider inputs. The current preview queues jobs visually until provider completion is enabled.
Can generated results be reused?
Once provider completion is enabled, finished generations can appear in history with preview, download, reuse, and copy actions.