GitHub - yasuomang/ai-video-prompt-cookbook: Practical prompt patterns for image-to-video and text-to-video workflows. · GitHub
/" data-turbo-transient="true" />
Skip to content
Search or jump to...
Search code, repositories, users, issues, pull requests...
-->
Search
Clear
Search syntax tips
Provide feedback
--><br>We read every piece of feedback, and take your input very seriously.
Include my email address so I can be contacted
Cancel
Submit feedback
Saved searches
Use saved searches to filter your results more quickly
-->
Name
Query
To see all available qualifiers, see our documentation.
Cancel
Create saved search
Sign in
/;ref_cta:Sign up;ref_loc:header logged out"}"<br>Sign up
Appearance settings
Resetting focus
You signed in with another tab or window. Reload to refresh your session.<br>You signed out in another tab or window. Reload to refresh your session.<br>You switched accounts on another tab or window. Reload to refresh your session.
Dismiss alert
{{ message }}
yasuomang
ai-video-prompt-cookbook
Public
Notifications<br>You must be signed in to change notification settings
Fork
Star
main
BranchesTags
Go to file
CodeOpen more actions menu
Folders and files<br>NameNameLast commit message<br>Last commit date<br>Latest commit
History<br>3 Commits<br>3 Commits
evaluations
evaluations
examples
examples
prompts
prompts
LICENSE
LICENSE
README.md
README.md
View all files
Repository files navigation
AI Video Prompt Cookbook
Practical prompt patterns for creators testing image-to-video and text-to-video workflows.
This cookbook is for creators, marketers, and small content teams who need usable AI video clips, not one-off demo prompts. It focuses on source image prep, motion wording, preservation constraints, repeatable testing, and failure review.
Who This Is For
Product marketers turning one product image into a short ad clip.
Social creators testing hooks, UGC-style motion, and vertical framing.
Brand teams that need a repeatable way to compare AI video outputs.
Editors who need clips that can still be cropped, captioned, and reused.
This is not an API integration guide. The goal is to make prompt testing easier for people who judge the output visually.
Prompt Card Format
Use one prompt card per test so the idea, constraints, and result stay connected.
Clip job:<br>Source image:<br>Subject to preserve:<br>Motion:<br>Camera:<br>Framing:<br>Style:<br>Negative constraints:<br>Success criteria:<br>Result notes:<br>Next change:
The two most important fields are Subject to preserve and Motion. Many weak AI video prompts describe the image again, but they do not explain what should move.
Image-to-Video Product Prompt
Use this when the source image is already the creative anchor.
Clip job: 5-second vertical product ad.<br>Source image: one clear studio image of the product.<br>Subject to preserve: keep the product shape, cap, label, color, and position unchanged.<br>Motion: soft light sweeps across the product surface while small background shadows move naturally.<br>Camera: slow push-in with stable framing.<br>Framing: 9:16 vertical, leave clean space near the top for a caption.<br>Style: realistic ecommerce product video, polished but not over-stylized.<br>Negative constraints: no extra objects, no label distortion, no melted edges, no background replacement.<br>Success criteria: product stays recognizable and the clip can be used in a short ad edit.
UGC-Style Hook Prompt
Use this when the output needs to feel like a creator shot, but the product still needs to stay readable.
Clip job: creator-style social hook for a product recommendation.<br>Source image: product held near a simple tabletop background.<br>Subject to preserve: keep the product label, package shape, and hand position stable.<br>Motion: subtle handheld movement, product tilts slightly toward camera, natural light shifts.<br>Camera: close vertical phone shot, mild handheld energy, no dramatic zoom.<br>Framing: 9:16, product centered, room for captions on the lower third.<br>Style: clean UGC product clip, realistic, casual, not cinematic.<br>Negative constraints: no fake text overlays, no changed packaging, no extra hands, no warped fingers.<br>Success criteria: the product is readable in the first second and can support a voiceover.
Same-Prompt Model Test
When comparing models, keep the job and prompt stable. Do not change the prompt after seeing the first result, or the comparison becomes a prompt rewrite test instead of a model comparison.
Test name: Product bottle, slow push-in, vertical ad.<br>Input: same source image for every model.<br>Prompt version: v1, unchanged for the first round.<br>Models tested: Model A, Model B, Model C.<br>Scoring: subject fidelity, motion coherence, prompt adherence, editability, retry cost.<br>Decision: keep, retry with constraints, or reject for this job.
Evaluation Scorecard
Criterion<br>Weight<br>What to Check
Subject fidelity<br>30%<br>Product, character, or object stays recognizable
Motion usefulness<br>20%<br>Motion starts early and supports the clip job
Prompt...