FACEHUB AI
ServicesUse CasesPricing
Sign In
Video

Bring Your Own Motion

Hand the model a clip and it follows the movement and the timing. Hand it photographs and it keeps the faces. Describe it alone and it invents everything — the choice is yours, shot by shot.
Filmmakers
Brands
Music Video Directors
Content Creators
Agencies
Try It Free View Pricing
FaceHub video composer with three reference photographs and one reference clip attached, a written shot description, and duration, aspect ratio and resolution set

What you hand it: your references, your shot, your timing.

A frame from the generated video: a woman and a man standing in a busy cafe, shot wide through the room

What comes back.

The Problem

Text-to-video invents the motion. You describe a slow push toward a doorway and get a drifting handheld wander; you describe a pause and get continuous movement. Stills fix the composition of one frame and say nothing about what happens next. So the shot in your head and the shot on screen keep failing to meet, and each attempt costs another generation.

The Solution

Attach what you already have. A reference clip carries the camera move and its timing, so the generated video follows a rhythm you set rather than one it chose. Reference images carry identity — a face, a product, a location — so the people stay themselves. Attach both and the model has the motion and the look; attach nothing and it works from your description alone.

How It Works

1
Attach your references

A clip for motion, images for faces and places, or a single still to animate. What you attach decides how the shot is made — you never pick a mode.

2
Write the shot

Say what happens, and name your references directly so the model knows who is who. Physical description of performance works far better than emotion words.

3
Choose length and quality, then generate

Five to fifteen seconds, standard or high. The cost is shown before you commit, and reference clips are billed by the second so you can see what a longer take costs.

What You Get

A reference clip sets the camera move and the timing

Reference images hold faces, products and locations steady

Animate a single still as an opening frame

Or generate from a description alone

Up to 15 seconds, standard or high quality

Cost shown before you generate, never after

Name references in the prompt to say who is who

Blocked scenes from the Scene Designer come straight through

Example Results

Generated video frame: a wide shot of a cafe, a woman in a dark coat facing a man across the room

The wide, as it was blocked.

Generated video frame: a close-up of the man at a cafe table, warm window light behind him

The cut to his close-up, on the beat it was given.

Generated video frame: a two-shot of the woman and the man, looking at each other across the cafe table

The two-shot — both of them still looking at each other.

Ready to Try It?

Start with 50 free credits. No credit card required.

Get Started Free

Frequently Asked Questions

Movement and timing — how the camera travels, when it stops, where the cuts fall. It does not control appearance: every surface in it is treated as a placeholder, and the look comes from your reference images and your prompt.

They do different jobs. Images fix identity — the same face, the same product. A clip fixes motion. For a shot where the camera has to do something specific, only a clip can say so; for a shot where a person must be recognisable, only images can.

Name them in the prompt. Write "Image 1 is the man in the charcoal coat" and the citation is rewritten into whatever form the underlying model expects before it is sent.

Between five and fifteen seconds per generation, which is the limit of the underlying models. Longer pieces are made by generating shots and cutting them together.

Credits scale with length and quality, and reference clips add a per-second charge because they are metered separately by the model. Reference images cost nothing extra. The full figure is shown before you generate.

Any video you own works. Or build one: the Scene Designer records your blocked scene — camera moves, cuts and all — as a clip made for exactly this.