Skip to main content
Take an image you’ve generated in Runchat and turn it into a short video, a flythrough, a style transfer, an exploded axonometric, a product shot. There are two main approaches and they solve different problems. Time: 5 minutes You’ll need: One image (for image-to-video) or two images (for first/last frame)

Open the example workflow

Make a copy of this workflow in your account to follow along.

​
Image-to-video for atmospheric motion

The basic approach: take an image, pick a video model, write a simple prompt describing the motion, run. Good for:
  • Drone pans
  • Slow reveals
  • Atmospheric motion across a landscape
The model invents whatever surrounds and follows the frame. Prompt patterns that work: 
Drone pan upwards to the horizon.

Slow camera push in towards the building.

Sweep across the landscape.
Keep it short and describe motion, not measurements. The output is usually 5 to 8 seconds. The trade-off: you have no control over what appears in the rest of the frame. The model fills it in. If you need specific geometry or framing, image-to-video alone isn’t enough.

​
First/last frame for control

When you want control over where the shot ends, use a first/last frame model. You provide two images and the model interpolates the motion between them. The workflow:
  1. Pick your starting image
  2. Pick your ending image. Often this is the same subject from a different angle, in a different style, or in a different state (e.g. exploded axonometric)
  3. Add a prompt describing the motion: The camera slides sideways to the right at a steady pace, keeping the building in view.
  4. Run
Useful patterns:
  • Same building, different angles: architectural flythroughs
  • Real render to exploded axonometric: design explainers
  • Product shot to detail shot: product marketing
  • Two finished renders, same subject: style transitions

​
Choosing a model

Runchat ships with several video models at very different price points. Cheaper models work for simple motion and short clips. More expensive models produce noticeably better results, especially for first/last frame where the model has to interpolate complex geometry. On cheaper first/last frame models, you’ll often see a cut in the middle, or aggressive morphing of the subject to force one frame into another rather than a smooth blend. Higher-quality models handle the interpolation more cleanly. Video is the most expensive generation in Runchat. Start with cheaper models to test your prompt and framing, then upgrade for the final version. See Credits for per-second pricing.

​
Combining clips

Runchat can combine multiple video outputs into a single clip without leaving the canvas. Useful for:
  • Stitching a flythrough with a detail shot
  • Assembling a sequence of first/last frame clips into a project reel
Add a Combine Videos node, connect your clips in order, run.

​
Next steps