Best Practices

Get the most out of Ozor.

A practical guide to creating great AI videos — from writing your first prompt to exporting at 4K. Whether you're new or a returning user, these principles will help you get more out of every credit.

How it works

01

Describe your video

Type what you want in plain English — the topic, style, audience, and structure. No video editing skills required. The more detail you give, the better the result.

02

AI generates scenes

Ozor's AI agent builds each scene as an animated React component, rendering at 30 FPS with real-time preview. You can edit, regenerate, or reorder any scene mid-flow.

03

Export your video

When you're happy, export to MP4 at 720p, 1080p, or 4K depending on your plan. The file downloads automatically — ready to publish anywhere.

Core principles

Prompting

Write better prompts

  • State the purpose up front — "a 30-second product launch video for our new iOS app"
  • Describe the visual style — "clean, minimal, dark background with blue accents"
  • Specify the audience — "aimed at developers" or "for non-technical executives"
  • Mention the tone — "confident and punchy" or "warm and conversational"
  • Include a word count or duration hint — "keep it under 60 seconds" or "aim for 4 scenes"
Scenes

Structure your scenes

  • Think in scenes: intro → problem → solution → CTA works for almost any video
  • Ask for a specific number of scenes — "create 4 scenes" gives more predictable results
  • Edit one scene at a time — click a scene in the editor before asking for changes
  • Request specific transitions — "slide in from left" or "fade out slowly"
  • Give each scene a clear job: hook, explain, prove, or close
Assets

Use your assets well

  • Upload brand assets before starting — logo, product screenshots, brand colors
  • Attach images directly to your chat message to include them in the next scene
  • Reference uploaded assets by name — "use the product screenshot I uploaded"
  • Set up your Brand Kit once and all future videos will use it automatically
  • Use the Library for stock animations and audio that are already optimized for video
Iteration

Iterate like a pro

  • Start broad, then refine — get the structure right before tweaking details
  • Use "make scene 2 shorter" or "make the text larger in scene 1" for precise edits
  • Regenerate a single scene without rebuilding the whole video
  • Ask for alternatives — "give me 3 different headline options for this scene"
  • Save scenes you love as templates so you can reuse them in future projects

Choosing your format

Ozor supports landscape (16:9) and portrait (9:16). Pick before you start — changing aspect ratio later requires regenerating your scenes.

Landscape — 16:9

1920 × 1080 px

The default. Best for presentations, YouTube, LinkedIn, website embeds, and any screen-forward context.

  • YouTube & Vimeo
  • LinkedIn video posts
  • Website hero / landing pages
  • Investor decks & presentations
  • Webinar & conference screens

Portrait — 9:16

1080 × 1920 px

Optimized for mobile-first platforms. Full-screen vertical video performs significantly better on short-form channels.

  • Instagram Reels & Stories
  • TikTok
  • YouTube Shorts
  • Pinterest video pins
  • Snapchat & BeReal

Common use cases

These are the video types that work best with Ozor. Each one includes recommended format, scene count, and a quick pro tip.

Product launches

Announce a new product, feature, or update. Show the before/after, highlight the key benefit, and close with a CTA.

Landscape (16:9)4–5 scenes

Lead with the outcome, not the feature list.

Social ads

Short, punchy videos optimized for Instagram Reels, TikTok, or YouTube Shorts. Hook viewers in the first 2 seconds.

Portrait (9:16)2–3 scenes

Put the hook in the first scene — no slow introductions.

Explainer videos

Walk through a concept, process, or product step by step. Great for onboarding, help centers, and sales.

Landscape (16:9)5–7 scenes

Use the problem → solution → proof structure.

Release notes

Ship a short video alongside your changelog. Each scene covers one new feature or improvement.

Landscape (16:9)1 scene per feature

Keep each scene to a single idea — no walls of text.

Investor updates

Summary of metrics, milestones, and roadmap. Professional style, data-driven, designed for busy founders.

Landscape (16:9)5–6 scenes

Mention specific numbers — percentages and round figures read well on screen.

Tutorials

Step-by-step walkthroughs with annotated screenshots or screen recordings layered into the video.

Landscape (16:9)1 scene per step

Number each step clearly and use "Next:" transitions between scenes.

Prompt formulas

Copy and fill in these templates to get started quickly. Replace the [brackets] with your own details.

Product launch

Create a [duration] product launch video for [product name].
Scenes: (1) Hook — one-line problem statement, (2) Solution — show the product in action, (3) Key benefits — 3 bullet points, (4) CTA — [your call to action].
Style: [minimal / bold / dark / light]. Brand colors: [colors]. Target audience: [who].

Social short

Create a 15-second vertical video (9:16) for [platform].
Scene 1: Bold hook — "[attention-grabbing statement]".
Scene 2: Proof or punchline — [what makes it believable or interesting].
Style: [energetic / calm / funny]. Use large text and minimal background.

Feature announcement

Make a [duration] video announcing [feature name] for [product].
Explain the problem it solves in one sentence. Show how it works visually. Close with a short CTA.
Keep the tone [professional / casual / excited]. Use [brand color] as the primary accent.

Explainer / How it works

Create a step-by-step explainer for [topic or process].
Break it into [N] scenes, one step per scene. Number each step clearly.
Audience: [who this is for]. Tone: [simple and clear / technical / conversational].
Use a [light / dark] background with [color] accents.

Prompt examples

The difference between a mediocre and a great AI video is almost always the quality of the prompt.

Too vague

Make a video about our product

No style, no audience, no content direction. The AI will guess everything and likely produce something generic.

Much better

Create a 45-second launch video for our project management app. 3 scenes: (1) the problem of disorganized teams, (2) how our app solves it with a live dashboard, (3) a CTA to try it free. Use a dark, professional style with white text and blue accents. Target audience: startup founders.

Clear duration, scene structure, visuals, tone, and audience. The AI has everything it needs.

Common mistakes

Things that waste credits and lead to frustrating results — and how to avoid them.

Being too vague

Fix: Instead of "make a cool video," say "make a 30-second dark-themed launch video for a SaaS app, targeting startup founders, with 3 scenes: problem, solution, CTA."

Describing what you want changed without targeting a scene

Fix: Click the scene thumbnail in the editor first, then ask for changes. This keeps edits surgical instead of regenerating the whole video.

Uploading assets after writing the prompt

Fix: Upload images and brand assets first, then reference them in your message. The AI can only use what's already available.

Requesting too many changes at once

Fix: One change per message works best. Ask for the text first, approve it, then refine the animation. Stacked requests lead to mixed results.

Skipping the Brand Kit

Fix: Set your Brand Kit once — logo, colors, font preferences, tone of voice. Every subsequent video will inherit these automatically.

Choosing the wrong aspect ratio

Fix: Pick 16:9 for presentations, YouTube, and LinkedIn. Pick 9:16 for Instagram Reels, TikTok, and YouTube Shorts. Changing after the fact requires regenerating all scenes.

Export guide

Video exports automatically download to your device once complete. Choose your resolution based on where the video will be published.

QualityAvailable onBest for
720pFreeDrafts, previews, internal sharing
1080pProSocial media, landing pages, presentations
4KBusinessConference screens, broadcast, premium campaigns

Exports are rendered server-side and typically complete within 2–5 minutes depending on video length and complexity. You'll see a progress indicator in the bottom-right corner of the editor.

Quick reference — editor commands

"Edit scene 2 to..."Target a specific scene for changes
"Add a new scene after scene 3"Insert a scene at a given position
"Make the text larger in scene 1"Adjust visual details in a single scene
"Change the color palette to..."Restyle the whole video at once
"Use the logo I uploaded"Pull in an asset from your library
"Give me 3 headline options"Ask for alternatives before committing
"Make this scene shorter"Trim the duration of a specific scene
"Add background music"Open the music picker for the current project
"Save this as a template"Reuse a scene or project structure later
"Export at 1080p"Trigger export with a specific resolution
"Regenerate scene 2"Redo one scene without touching the rest
"Change the aspect ratio to 9:16"Switch to vertical / portrait format