What Runway is
Runway is an AI video suite rather than a single generator. Generation, editing, effects and collaboration live in the same interface, which is the main reason teams choose it over a pure model API.
- Text to video and image to video generation.
- A timeline editor so you can cut generated clips together without leaving the tool.
- Camera and motion controls for directing movement precisely.
- Effects and restyling applied to existing footage.
- Team workspaces for shared projects.
Who it suits: teams that want one tool from generation to cut. If you already edit in Premiere or Resolve, the generation quality matters more than the built-in editor.
Getting set up
-
Create a workspace
Projects live in a workspace. Set one up per client or per brand so assets do not blur together.
-
Understand credits before generating
Generation consumes credits. Check current rates and set an internal budget before your first session.
-
Decide your delivery resolution up front
Generating at final resolution costs more. Decide early whether you will refine later.
-
Prepare input assets
If using image-to-video, gather clean reference images first.
Generating from text
Text to video is best for concepting and for footage that does not exist yet. Write prompts that specify motion and camera, not just subject.
- State what changes during the shot — that is what makes it a video prompt.
- Name one camera move, or say the camera is static.
- Keep it to two or three sentences.
Generating from an image
Use this whenever you already know what the frame should look like. It is more predictable than text to video and is the right default for product and branded work.
- Upload a clean, well-lit source image.
- Describe only the motion you want, not the whole scene — the image already carries the visual.
- Keep motion simple in the first pass and add complexity only if needed.
- Generate several variations and pick rather than trying to hit it in one attempt.
The default for real work: if you have a reference image, use it. Image-to-video preserves what you care about; text-to-video re-invents it every time.
Directing the camera
| Control | Use for | Caution |
|---|---|---|
| Pan | Revealing a scene sideways | Keep speed slow |
| Tilt | Revealing vertically | Slow speed |
| Push in / pull out | Emphasis or reveal | Most reliable single-axis move |
| Orbit | Showing an object all round | Can distort the subject |
| Custom motion path | Specific directed movement | Complex paths increase artifacts |
One move per shot. Combined moves are where subject distortion and flicker appear.
Editing on the timeline
-
Generate more clips than you need
Approval rate is never 100 percent; over-generate deliberately.
-
Assemble a rough cut
Get structure right before polishing anything.
-
Trim to the strongest part of each clip
Generated clips usually have a good middle and weak ends.
-
Colour match between clips
Generated shots rarely match; grade them to sit together.
-
Add audio last
Music, voiceover and effects should be balanced after the picture is locked.
Managing credits
- Approve at lower resolution, then spend on final output only for approved clips.
- Do not batch-generate at maximum settings — most output will be discarded.
- Keep prompts and seeds so approved work can be reproduced rather than re-guessed.
- Track your own approval rate; it is the real driver of cost, not the credit price.
Mistakes that waste credits
| Mistake | Better approach |
|---|---|
| Generating long clips hoping for a usable section | Generate short clips and cut them together |
| Specifying several camera moves at once | One move per shot |
| Trying to generate readable text | Add text as an overlay in the editor |
| Generating at final resolution every time | Approve low, then refine approved clips |
| Rewriting prompts from scratch each time | Build templates and vary one element |