Still to Motion: Building One Two-Step AI Video Asset From Brief to Brand Layer

Marketing Academy · Field Work●AI in Marketing

All projects·From: AI Image & Video Creation

CoreBuild the Asset· 75 minutes

Coinbase

Objective: Run the lesson's full playbook on one asset: write a structured 6-element prompt, generate a still image, animate it with image-to-video, then apply the brand layer before export.

You're the creative producer at Coinbase, the Nasdaq-listed (COIN) crypto exchange, building one 8-second social clip for a new feature launch with no video shoot budget.

Write the brief, write a structured prompt, generate and evaluate a still, animate it, and pass it through the brand layer before it's export-ready.

Before you start

What you'll need

Free path (everything below is enough to finish)

Google Sheets

FreeTrack prompt versions and checklist status

Free, and keeps a running prompt library so a working prompt is never lost

Canva

FreemiumApply the brand layer: hex codes, logo, and disclaimer text

Free tier includes brand kit and text overlay tools

Paid upgrades (optional, faster/deeper)

The full free path is DALL-E 3 via ChatGPT's free tier for the still plus Canva's free tier for the brand layer; Midjourney and Runway are quality upgrades, not requirements, for this exercise.

Midjourney(optional)

PaidGenerate the structured-prompt still image

Highest photorealistic quality for the still-image step

No access? DALL-E 3 via ChatGPT's free tier produces a usable still for this exercise

Runway(optional)

FreemiumImage-to-video animation of the approved still

Gen-3's image-to-video mode is the lesson's recommended controllable-motion tool

No access? Kling or Pika's free trial credits cover a single 8-second clip

The process

3 steps

Step 01 of 03

Structured 6-element prompt writing

A strong prompt specifies subject, setting, lighting, style, mood, and format; a weak prompt like 'a marketing image for a crypto app' produces a generic result.

Your brief is: a feature-launch hero image for a mobile trading app, Instagram feed, calm and trustworthy mood. Write the 6-element prompt.

Midjourney— Midjourney's prompt box, or DALL-E 3 if you don't have Midjourney access.

Procedure

Name the subject: who or what
Name the setting: where
Specify lighting
Specify style, such as photorealistic or illustration
Specify mood
Specify format and aspect ratio for the target platform

Sample output

PROMPT v1
'a calm young professional checking a trading app on their phone, minimalist home office, soft afternoon window light, photorealistic, 35mm film, trustworthy and calm mood, 4:5 aspect ratio for Instagram feed'

WEAK PROMPT REJECTED: 'a marketing image for a crypto app'

Healthy

The prompt names all 6 elements explicitly; the output matches the brief on the first or second generation.

Unhealthy

A one-line vague prompt requires 8+ regenerations because nothing was specified.

What this means

Every unspecified element is a coin flip the model makes for you; specify it or accept the randomness.

So what do I do about it?

Symptom	Action	Effort
Regenerating the same vague prompt 10+ times	Stop and add the missing elements instead of re-rolling	5 min

YouYou can do this yourself, no engineering access required.

Step 02 of 03

Two-step image-to-video workflow

Text-to-video is unpredictable for brand-controlled content; generating a still first, then animating it in Runway or Kling, keeps composition and color under your control and only adds motion in the second step.

Your still image passed review. What's the minimum motion to add for an 8-second social clip without introducing new errors?

Runway— Runway Gen-3's image-to-video tool, using the approved still as the input frame.

Procedure

Import the approved still into Runway's image-to-video tool
Choose one motion type only, such as a slow zoom, parallax shift, or pan, not multiple stacked effects
Cap the clip at 8-10 seconds for social use
Review the output at 100% for morphing or distortion before moving on

Sample output

CLIP SPEC
  Input: approved still (4:5)
  Motion: slow zoom, 1.05x over 8s
  Output: 8s, 1080x1350, MP4
  Review: no morphing detected on 2 passes

Healthy

One clean motion effect, under 10 seconds, with no new visual errors introduced.

Unhealthy

A 20-second clip with 3 stacked motion effects introduces a warped hand mid-clip.

What this means

Every additional second and every stacked effect is another chance for the model to introduce an error; minimal motion on a pre-approved still is the safer path.

So what do I do about it?

Symptom	Action	Effort
Text-to-video output has a morphing face or object	Fall back to the two-step method: still first, then minimal animation	30 min

YouYou can do this yourself, no engineering access required.

Step 03 of 03

Brand layer application

AI tools don't know your exact hex codes, logo placement rules, or legal copy; every output must pass through a brand template before publish.

The animated clip is otherwise clean. What has to be added before it's export-ready for a public feature launch?

Canva— Canva's brand kit, importing the Runway output as a layer.

Procedure

Overlay the exact brand hex codes via the color-grade or brand kit tool
Place the logo per the brand's placement and minimum-opacity rules
Add any required legal disclaimer text for a financial product
Export at the platform's required resolution and aspect ratio

Sample output

BRAND LAYER CHECKLIST
  [x] Hex codes matched to brand kit
  [x] Logo placed bottom-right, 100% opacity, no overlap
  [x] Disclaimer text added: 'Crypto asset values can go up or down.'
  [x] Exported 1080x1350 MP4 for Instagram feed

Healthy

Every brand-layer item is checked before export; the clip reads as unmistakably the brand's own.

Unhealthy

The raw Runway export is published directly, missing the legal disclaimer required for a financial product.

What this means

The brand layer is not optional polish; for a regulated product category it includes compliance text, not just aesthetics.

So what do I do about it?

Symptom	Action	Effort
AI output looks great but has no required disclaimer	Route every financial-product asset through legal or compliance before the brand layer is called done	30 min

EitherYou or a developer can handle this, depending on your access.

Final deliverable

One export-ready 8-second social video clip with a documented structured prompt, an approved still image, and a completed brand-layer checklist.

See a reference example

Sample output

Snowflake, feature-launch clip spec (excerpt)

PROMPT: 'a data engineer reviewing a live dashboard, modern office, cool blue accent light, photorealistic, calm focused mood, 16:9'
STILL: approved on generation 3
MOTION: slow pan, 8s
BRAND LAYER: hex #29B5E8 matched, logo top-left, no disclaimer required for a non-financial product

Success criteria

You're done when you can:

Prompt specifies all 6 required elements
Still image passes the anatomy, text, color, and logo checklist before animation
Clip stays under 10 seconds with a single motion effect
Brand layer checklist is fully completed before export