Runway and HeyGen Prompts: 10 Templates for Marketing Videos

Runway prompts should describe the shot. HeyGen prompts should describe the production. Use the templates below as reusable briefs: replace the words in brackets, generate one test, and change only the part that failed.

If you are still deciding which platform fits the job, start with our Runway vs HeyGen comparison for marketing teams. This guide assumes you already have a tool and need a faster way to get useful output.

Runway

Direct the shot

1. Frame and subject
2. Action and environment
3. Camera and timing
4. Light and visual style

What are you controlling? A shot or a complete video

HeyGen

Brief the production

1. Audience and goal
2. Exact script
3. Scenes and media
4. Brand, voice and format

Before you copy a prompt

Prepare five inputs first:

1. One deliverable: a six-second product shot, a 30-second ad, or a 60-second explainer.
2. One audience: name the buyer, user, or viewer.
3. One message: decide what the viewer should understand or do.
4. Reference assets: product images, screenshots, logos, brand colors, or a finished script.
5. Output rules: aspect ratio, duration, language, and channel.

Do not ask one generation to solve the strategy, script, art direction, and final edit at the same time. A narrow brief is easier to test and easier to repair.

Five Runway prompt templates

For text-to-video, describe both the visible scene and its motion. For image-to-video, the reference image already defines the subject and composition, so the prompt should focus mainly on movement.

1. Product hero shot

Use this for a landing page, launch film, or product reveal.

[SHOT SIZE] shot of [PRODUCT]
on [SURFACE OR SETTING].
[PRODUCT ACTION OR REVEAL].
The camera [CAMERA MOVEMENT]
while [ENVIRONMENTAL MOTION].
[LIGHTING]. [VISUAL STYLE].
The movement is [SPEED AND FEEL].

Example variables: macro close-up, a black running shoe, a wet stone platform, slowly rotates, pushes forward, fine mist moves through the backlight.

Improve it: if the product shape keeps changing, create or upload a strong reference image and switch to image-to-video.

2. Image-to-video product motion

Use this when the first frame is already approved.

The camera [CAMERA MOVEMENT]
as the product [SUBJECT MOTION].
[SECONDARY ELEMENT] moves
[DIRECTION AND SPEED].
[LIGHT OR SHADOW CHANGE].
Continuous seamless shot.

Do not repeat every detail visible in the image. Describe what should move, how it should move, and what the camera should do.

Improve it: if the frame drifts, ask for a motionless camera in positive language and give the subject a small, clear action.

3. Vertical social hook

Use this for the opening shot of a short ad.

Vertical [SHOT SIZE] of [SUBJECT]
in [SPECIFIC ENVIRONMENT].
In the first second, [IMMEDIATE ACTION].
The camera [FAST CAMERA ACTION],
revealing [PRODUCT OR RESULT].
Energetic movement and clear separation.
[BRAND COLOR] accents.
Designed as a strong opening shot.

Improve it: generate the hook as its own short clip. Add the offer, captions, and call to action in the edit instead of asking the video model to solve the whole ad.

4. Lifestyle B-roll

Use this to support an interview, avatar, voice-over, or product explanation.

[CAMERA STYLE] shot of [PERSON OR TEAM]
[CONCRETE ACTION] in [REAL ENVIRONMENT].
The camera [MOVEMENT].
Natural movement in the background.
[TIME OF DAY] light.
[MOOD] atmosphere.
Realistic commercial footage.

Replace vague words such as “professional” with visible detail: clean desk, neutral wardrobe, soft window light, focused expression, or an active warehouse floor.

Improve it: create a shot list before generating. Five precise B-roll prompts are more useful than one prompt asking for a complete story.

5. Timed action sequence

Use this when one short shot must contain a clear order of events.

[FULL VISUAL DESCRIPTION].
[CAMERA AND MOTION DESCRIPTION].
[00:00–00:02] [FIRST ACTION]
[00:02–00:04] [SECOND ACTION]
[00:04–00:06] [FINAL ACTION OR REVEAL]

Give each action enough time to happen. If the model skips a step, simplify the sequence or split it into separate shots.

Improve it: keep the natural-language description above the timestamps. The timestamps guide the order; they do not replace a clear description of the scene.

Five HeyGen prompt templates

HeyGen Video Agent works best when the prompt behaves like a production brief. Include the purpose, audience, script, scene direction, media type, and output format. If exact wording matters, provide the finished script instead of asking the tool to invent it.

1. Presenter-led explainer

Create a [DURATION] [FORMAT] explainer
for [AUDIENCE].
Goal: help the viewer understand
[ONE TOPIC] and take [ONE ACTION].
Use a [PRESENTER DESCRIPTION]
with a [VOICE AND TONE].
Use the exact script below. Do not add claims.

SCRIPT:
[PASTE APPROVED SCRIPT]

Visual direction:
[BACKGROUND, BRAND COLORS, TEXT STYLE].
Use short on-screen headings
for [KEY POINTS].
End with: [CALL TO ACTION].

Improve it: preview the voice before the final render. Fix pacing, punctuation, pauses, names, and acronyms while changes are still cheap.

2. 30-second vertical ad

Create a 30-second vertical video ad
for [AUDIENCE] on [CHANNEL].
Message: [ONE PROMISE].
Tone: [THREE SPECIFIC ADJECTIVES].

Scene 1 — Hook, 0–4 seconds
Visual: [AVATAR OR MEDIA]
Script: "[HOOK]"

Scene 2 — Problem, 4–11 seconds
Visual:
[STOCK, SCREENSHOT OR MOTION GRAPHIC]
Script: "[PROBLEM]"

Scene 3 — Solution, 11–24 seconds
Visual: [PRODUCT DEMO OR AVATAR]
Script: "[SOLUTION AND PROOF]"

Scene 4 — CTA, 24–30 seconds
Visual: [END CARD]
Script: "[CALL TO ACTION]"

Improve it: assign media by function. Use real screenshots for product truth, motion graphics for data, stock for real environments, and generated visuals for concepts that do not exist.

3. Product demo from screenshots

Create a [DURATION] product walkthrough
for [AUDIENCE].
Use the attached screenshots
in the order listed below.
Do not redesign the interface.

1. [FILE NAME]
   Show while explaining [FEATURE]
2. [FILE NAME]
   Zoom toward [CONTROL OR RESULT]
3. [FILE NAME]
   Show while explaining [OUTCOME]

Use this exact voice-over:
[PASTE SCRIPT]

Add restrained callouts around
the relevant interface elements.
Use [BRAND COLORS], clean transitions,
and a [VOICE STYLE] narrator.

Improve it: tell Video Agent why each attachment exists. A folder of unlabeled assets is not a production brief.

4. Localized campaign version

Create a [LANGUAGE] version
of this [FORMAT] video for [REGION].
Keep product names, plan names,
URLs, and technical terms unchanged.
Use the attached brand glossary
for pronunciation and translation rules.
Adapt idioms for natural [LANGUAGE].
Do not change the offer or factual claims.
Use [AVATAR], [VOICE], and [BRAND SYSTEM].
Final call to action: [LOCAL CTA].

SOURCE SCRIPT:
[PASTE APPROVED SOURCE SCRIPT]

Improve it: maintain a separate reviewed glossary for every production language. Check the script and voice before scaling the same version across a campaign.

5. One-scene batch variation

Use this for personalized outreach, short ad variants, or repeated updates.

Create one [ASPECT RATIO]
single-scene avatar video.
Audience segment: [SEGMENT].
Presenter: [AVATAR].
Voice: [VOICE].
Background: [SIMPLE BRAND SETTING].
Delivery: [PACE, ENERGY, EMOTION].

SCRIPT:
Hi [NAME].
[PERSONALIZED CONTEXT].
[ONE RELEVANT BENEFIT].
[LOW-FRICTION CTA].

Improve it: approve one complete sample before starting a batch. HeyGen Batch Mode is designed for repeated single-scene videos; use Studio when the story needs multiple scenes, B-roll, or transitions.

A faster prompt-to-video workflow

The biggest productivity gain does not come from making prompts longer. It comes from separating exploration, approval, and scale.

01 Lock the brief Audience, message, asset and format

→

02 Fill a template Use fixed fields, not a blank prompt box

→

03 Run one test One shot, scene, voice or audience

→

04 Change one field Keep successful variables fixed

→

05 Approve and scale Render the sequence, variants or languages

A controlled test makes failures easier to diagnose and prevents the same mistake from spreading through a batch.

Use these rules in daily production:

• Keep a prompt library. Save approved prompts by deliverable, not by vague campaign name: vertical-product-hook, avatar-feature-explainer, or localized-demo.
• Turn variables into fields. Mark [AUDIENCE], [SHOT], [SCRIPT], and [CTA] so another team member can reuse the prompt safely.
• Change one variable per iteration. If camera, lighting, action, and style all change together, you cannot tell what improved the result.
• Separate shots from edits. Generate clean source clips in Runway. Add typography, logos, legal copy, and exact timing in the editing layer.
• Approve scripts before rendering avatars. In HeyGen, a script or voice change can require a new render. Check pronunciation and pacing first.
• Name versions consistently. Include campaign, asset, market, ratio, language, and version: launch_hook_US_9x16_EN_v03.
• Scale only approved units. One approved shot can become a sequence; one approved avatar scene can become a batch or localization set.

Common prompt failures and fixes

Problem	Likely cause	First fix
Runway output feels random	The prompt describes a mood but not visible action	Add a concrete subject action and camera movement
Product identity drifts	Text-to-video has no stable product reference	Start from an approved product image
The shot contains unwanted movement	The camera instruction is missing or ambiguous	State that the camera remains motionless and define subject motion
HeyGen chooses weak visuals	The brief does not assign media to each scene	Specify screenshot, stock, motion graphic, avatar, or generated visual
The avatar sounds unnatural	Script punctuation and pronunciation were not reviewed	Preview the voice; add pauses and glossary rules
Batch output repeats an error	Production scaled before approval	Generate and approve one representative sample

Frequently asked questions

Should Runway prompts be long?

Only as long as needed to remove important ambiguity. Start with the subject, scene, motion, and camera. Add light, style, or timing when the first result shows what is missing. Long prompts with conflicting instructions are harder to troubleshoot.

Should I use JSON prompts in Runway?

You do not need JSON. A consistent natural-language structure is easier to read and iterate. The useful part is separating the subject, motion, camera, and style—not the brackets or syntax.

Should I let HeyGen write the script?

Use automatic scripting for a rough internal draft. For a public campaign, provide an approved script when wording, claims, pronunciation, or timing matters. This gives the production system less room to change the message.

Can one prompt create an entire campaign?

It can create a starting point, but reliable campaigns are modular. Approve the script, visual direction, representative shot, avatar voice, and CTA separately. Then reuse those approved parts across formats and markets.

The practical rule

Treat prompts as production templates, not one-off inspiration. Runway needs clear visual direction for each shot. HeyGen needs a structured brief for the whole video. When the team locks the variables, tests one small output, and scales only what passed review, video production becomes faster and far more predictable.

To budget the pilot beyond subscription prices, use our AI video cost framework and measure the cost of each approved asset.