Alien Road uses Sora and other AI video generation tools for concept films, social variants, backgrounds and B-roll in client campaigns, inside a process with written briefs, human review, clear labelling and side-by-side measurement against conventional video. The tools speed up production; they do not replace the decisions about what the brand should say.
Sora is OpenAI’s video generation model. It creates short clips from text and images and is available through OpenAI’s products, including a dedicated Sora app, with features and access that have changed over time. OpenAI adds provenance signals such as C2PA metadata and visible watermarks to Sora output in many cases. Other tools in the same category include Google Veo, Runway and Kling, and the right choice depends on the shot, the licence terms and where the client works.
Where does AI video fit in a campaign?
We sort video needs into three groups. Hero content, such as a launch film or TV spot, usually stays with conventional production, sometimes with AI shots inside it. Supporting content, such as social cut-downs, explainers and seasonal variants, is where AI video does the most work. Exploration, such as mood films and pitch concepts, is almost entirely AI now, because testing ideas is fast and cheap.
The sorting matters because it sets expectations. A client who sees an AI mood film knows it is a concept, not the final ad.
Step 1: the brief comes before the prompt
Every AI video starts from a normal creative brief: objective, audience, message, mandatory elements, tone and format. From the brief we write a shot list, and only then prompts. Skipping the brief produces pretty clips that do not sell anything.
- Objective and single key message.
- Platforms and aspect ratios: 9:16 for Reels and TikTok, 1:1 or 4:5 for feeds, 16:9 for YouTube.
- Mandatory real assets: logo, product pack shots, legal lines.
- Restricted content: real people, competitor references, claims.
- Visual references from the brand’s own library.
Step 2: generation and iteration
Prompts describe subject, setting, camera movement, lighting and mood in plain language. We keep a shared prompt library per client so successful styles can be reused. Each shot is generated several times, and the editor selects takes the way they would from a real shoot.
Known weak points are planned around. AI video often struggles with readable text, exact logos, hands and precise product detail. Real product footage and logos are composited in the edit, so the product on screen is always the actual product.
Step 3: review and approval
Each cut passes three checks before the client sees it. A creative check covers brand fit and quality. A truth check confirms that nothing shows the product doing what it cannot do and no scene implies a real event that did not happen. A rights check confirms no real person’s likeness, no imitation of a named artist or film, and music licensed separately.
The client then approves as with any video. We tell them which shots are generated, so approval is informed.
Step 4: labelling and platform rules
Realistic AI video is labelled according to the rules of the platform where it runs. Meta, TikTok and YouTube all have policies on AI-generated or altered content, and some apply labels automatically when provenance metadata is present. We keep metadata intact rather than stripping it, and add visible disclosure where the content could be mistaken for real footage.
For EU audiences, the AI Act’s transparency obligations for synthetic content add another reason to have a consistent labelling rule.
Step 5: measurement
AI video is measured the same way as other video, with the comparison built in. Where budgets allow, we run AI-produced variants against conventional ones in the same ad set or campaign and compare hook rate, hold rate, click-through and cost per result.
- Hook rate: share of impressions that watch the first seconds.
- Hold rate: share of viewers reaching a set point, such as the midpoint.
- Click-through and cost per click or lead.
- Production hours and cost per finished asset.
- Comments mentioning AI or authenticity, positive or negative.
What results should a client expect?
The clearest gain is volume and speed of variants, which helps paid social where creative fatigue is a constant problem. Quality gains depend on the shot type; abstract, atmospheric and stylised scenes come out strong, while close product work still needs real footage. We avoid promising performance uplift in advance. Some audiences respond well to AI visuals, others react against them, and only testing shows which applies.
Which mistakes do we avoid?
Generating before the brief is agreed. Showing a client an AI concept without saying it is a concept. Using AI for testimonial-style content, which is misleading by nature. Relying on one tool for everything when another handles a particular shot better. And letting the prompt library live in one person’s head instead of a shared document.
How Alien Road applies AI video in a typical engagement
In practice, discovery looks at the client’s existing video library and performance data to find where variants are most needed, usually paid social with frequent creative refreshes. Setup produces the brief template, a client-specific prompt library, reference image sets and the labelling rule. Production runs in short cycles: generate, edit, review, publish, measure. Every monthly report shows which generated assets ran, how they performed against conventional ones and which shots were rejected in review, so the client sees the full picture rather than the best clips only.
How do Sora and similar tools change the team?
Roles shift rather than disappear. Writers spend more time on shot lists and less on long treatments. Editors become selectors and compositors. Producers manage licences and provenance alongside schedules. The people who understand the brand matter more, because the tools will produce anything, and someone has to decide what is right.
Where to start this week
Take a video you already plan to make and write its brief and shot list as usual. Generate only the B-roll shots with Sora or a similar tool, composite your real product footage, and run the finished piece against last month’s conventional version. One honest comparison is a better starting point than a full switch.
Keep reading: Video Production · AI Marketing