Runway is a New York-based company whose web platform generates and edits video with AI, including text-to-video, image-to-video, motion controls and editing tools such as background removal. Moving from a pilot to production means setting clear use cases, a prompt and asset library, quality checks, rights rules and a review process. Without those, teams get impressive demos and few usable clips.
What is Runway good at today?
Runway’s Gen series of video models turns a prompt or a still image into a short clip, typically a few seconds long, which can be extended or combined. It also offers tools for camera movement control, reference-based consistency, upscaling and removing or replacing objects. Model generations and features change quickly, so check the current model list in the app.
In marketing, the strongest uses are short: B-roll for social ads, animated product stills, mood films for pitches, background plates and quick visual tests before a shoot. Long narrative pieces with consistent characters and dialogue remain hard, and most teams still combine Runway clips with real footage in a traditional editor.
Phase 1: How should you scope the pilot?
A pilot should answer one question: can this tool produce clips we would actually publish, faster or cheaper than today? Pick a narrow use case with a real deadline. ‘Explore AI video’ is not a pilot. ‘Produce six animated product shots for next month’s Instagram campaign’ is.
- Pick one use case and one channel
- Set a success bar: for example four of six clips approved without reshoots
- Choose two to three people: a video editor, a designer and a reviewer
- Set a credit budget for the pilot and track usage in the account settings
- Define the timeline: two to four weeks
- Collect the baseline: current cost and time for the same output
Phase 2: Inputs and the prompt library
Runway output improves sharply with good inputs. Image-to-video is usually more controllable than text-to-video for brand work, because you start from an approved still. Feed it a clean product photo or a designed key visual rather than hoping the model invents your product correctly.
Save every prompt that works, with the input image, settings and the clip it produced. Within a few weeks you will have a library of camera moves and styles that match the brand. New team members start from that library instead of from zero.
- Approved product stills at high resolution, cut out where needed
- Brand color references and lighting notes
- Camera movement presets: slow push-in, orbit, static with subject motion
- A list of things to avoid: logos warping, text in frame, hands holding products
- Aspect ratio standards per channel: 9:16, 1:1, 16:9
Phase 3: What quality checks are needed?
AI video has its own failure modes. Products can change shape between frames, labels can turn into nonsense letters and physics can look wrong. A reviewer who knows these patterns catches them quickly; someone seeing AI video for the first time may miss them until a customer points them out.
Frame-level QA checklist
- Product shape, color and proportions stay consistent through the clip
- No distorted text, logos or packaging details
- No extra fingers, limbs or faces in the background
- Motion looks plausible: liquids, fabric, hair
- No resemblance to real people or copyrighted characters
- Clip exported at the right resolution and frame rate for the channel
If a product detail must be exact, such as a regulated label, do not rely on generated pixels. Composite the real label over the clip in After Effects or a similar tool, or use real footage for that shot.
Phase 4: Rights, disclosure and approvals
Read Runway’s terms for your plan regarding ownership and commercial use, and check whether your contract includes any indemnity. Keep a record of each published clip’s prompt and inputs. For ads in the EU, consider the EU AI Act’s disclosure duties for realistic synthetic media, and follow each ad platform’s policies on AI-generated content.
Approvals should be explicit. Clients and legal teams are entitled to know which shots were generated. A simple column in the delivery sheet, ‘AI generated: yes or no’, avoids awkward questions later and builds trust.
Phase 5: Moving into production
Production readiness means the workflow runs without the original pilot team in the room. That requires documented steps, a shared asset folder, naming conventions and a clear owner for the account and credits. Runway’s team plans include shared workspaces and asset libraries, which help with handover.
- Document the workflow from brief to export in one page
- Set naming conventions: client_campaign_shot_version
- Store inputs, prompts and outputs together
- Assign a credit owner and a monthly budget alert
- Schedule a monthly review of new features and update the playbook
- Define when not to use Runway: hero films, testimonials, regulated claims
How do you measure whether it worked?
Compare the pilot output with the baseline on three measures: cost per approved clip, hours per approved clip and creative performance. For paid social, run AI-assisted clips against standard ones and compare hook rate, average watch time and cost per result over the same audience and budget. Be fair: test like with like.
How Alien Road applies Runway
In a typical engagement our video team runs the pilot alongside a live campaign, so the test is real. Discovery covers the client’s asset library, brand rules and channels. Setup includes the prompt library, QA checklist and delivery sheet with AI labeling. Every generated shot passes an editor and a reviewer before it reaches the client.
Reporting compares cost and time per approved clip with the client’s previous production method, plus ad performance where clips ran in paid media. If the numbers support it, we expand to more formats; if not, Runway stays a pre-production tool for storyboards and mood films, which is still useful.
How do you budget credits?
Runway plans run on credits, and video generation uses them quickly, especially at higher resolution and longer durations. During the pilot, log credits spent per approved clip, not per generation. That number is what matters for production planning, because many generations are discarded.
Draft at lower settings where the app allows it, then spend more credits only on the shots that work. Teams that generate every exploration at maximum quality burn through a month’s allowance in days.
Where to start this week
Choose one product photo you already own and approve. Upload it to Runway, generate five short clips with different camera moves, and review them against the QA checklist above. Count how many you would actually publish. That number is your first honest data point.
Keep reading: Video Production · AI Marketing