Sora 2 Prompt Guide: Timecode Tricks + Real Examples from Viral Videos

My first Sora 2 video was a disaster.

I typed "a cat playing with a toy in a living room" and hit generate. What came back looked nothing like what I imagined—wrong lighting, weird camera angles, and the cat morphed halfway through.

After studying hundreds of viral Sora 2 videos on X (Twitter), I discovered something: the best creators aren't just writing prompts. They're writing shot lists with timecodes.

Here's everything I learned about writing prompts that actually render what you want.

The Sora 2 Prompt Formula

Think of prompting like briefing a cinematographer who has never seen your storyboard. If you leave out details, they'll improvise—and you may not get what you envisioned.

OpenAI's official guide recommends structuring your prompt in layers:

Layer	What to Include	Example
Frame	Aspect ratio, duration	`16:9, 10 seconds`
Scene	Location, time, weather	`cozy living room, golden hour light`
Subject	Character, wardrobe, features	`small fluffy white cat`
Action	Movement, gestures, beats	`paws at a toy, then looks up`
Camera	Lens, movement, angle	`eye-level, slow push in`
Style	Mood, reference, quality	`cinematic, warm tones, 4K`

Timecode: The Secret Weapon

Here's what separates amateur prompts from viral ones: timecode notation.

Instead of describing everything at once, you break your video into shots:

(0–3s): [Shot 1 description]
(3–6s): [Shot 2 description]
(6–10s): [Shot 3 description]

This tells Sora exactly when to transition between scenes, camera angles, or actions.

Real Example: Viral Cat Video

This prompt from @BartzWow generated a video with over 100K views:

A small fluffy cat in a cozy, golden-hour living room paws a toy while
on-screen text reads 'What is this little guy doing?' (0–1.5s).
Camera at eye-level slowly pushes in as the cat behaves normally (1.5–3s).
Suddenly, the cat performs a surprising backflip (3–5s).

Why it works:

Clear timecode breaks: (0–1.5s), (1.5–3s), (3–5s)
Specific camera direction: "eye-level slowly pushes in"
Unexpected beat: "surprising backflip" creates the viral moment

Real Example: Product Unboxing

From the OpenAI Cookbook:

A cinematic unboxing of a premium mirrorless camera on a wooden table.

(0–3s): Slow dolly in from the right, shallow depth of field,
warm morning light through a window, dust motes visible.

(3–8s): Top-down 45° reveal as hands open the box,
soft foley of cardboard and magnetic clicks.

(8–12s): Cut to 3/4 profile of the camera on velvet cloth,
subtle lens flare, soft ambient synth pad.

Two Approaches: Simple vs. Detailed

The Minimalist Way

Sometimes a single sentence + reference image is enough:

Make a professional ad for [your product]

This works when:

You have a strong reference image
You want Sora to be creative
You're exploring ideas quickly

@CharaspowerAI proved this works—his simple prompt + product image generated professional-looking ads that went viral.

The Director's Way

For full control, write a complete shot list:

16:9, 12 seconds, cinematic product commercial.

(0–3s): Close-up of running shoes on wet pavement,
golden sunrise light, shallow DOF, brand logo visible.

(3–7s): Low-angle slow-motion of runner's feet hitting ground,
dust particles flying, dynamic motion blur.

(7–12s): Wide aerial shot following runner through city streets,
energetic mood, cinematic color grade.

5 Copy-Paste Templates

1. Product Advertisement

A cinematic commercial for [PRODUCT NAME], 16:9, 10 seconds.

(0–3s): Close-up product shot on minimal background,
soft spotlight, premium textures visible.

(3–7s): Product in use, lifestyle context,
natural lighting, aspirational mood.

(7–10s): Hero shot with brand logo,
slow zoom in, modern and impactful.

Style: Ultra-realistic, 4K, professional advertising quality.

2. Pet/Animal Video

[ANIMAL DESCRIPTION] in [SETTING], golden hour lighting.

(0–2s): Wide establishing shot, cozy atmosphere,
[animal] enters frame naturally.

(2–5s): Eye-level medium shot, camera slowly pushes in,
[animal] does [NORMAL ACTION].

(5–8s): Sudden [SURPRISING ACTION],
slight slow-motion for comedic effect.

Style: Warm, nostalgic, documentary feel.

3. Travel/Landscape

Breathtaking [LOCATION] at [TIME OF DAY], 16:9, 8 seconds.

(0–4s): Sweeping drone shot revealing landscape,
golden light, mist in valleys, epic scale.

(4–8s): Ground-level tracking shot through [terrain],
foreground elements passing by, immersive POV.

Style: Cinematic, National Geographic quality, vibrant colors.

4. Fashion/Lifestyle

[PERSON DESCRIPTION] in [SETTING], editorial fashion style.

(0–3s): Full body shot, model walking toward camera,
confident stride, clothes flowing naturally.

(3–6s): Close-up on outfit details,
shallow depth of field, texture visible.

(6–9s): Portrait shot, model looks at camera,
soft smile, backlit with golden rim light.

Style: High fashion editorial, Vogue aesthetic, 4K.

5. Food/Culinary

[DISH NAME] preparation and presentation, overhead angle.

(0–3s): Ingredients arranged on wooden surface,
hands enter frame, begin preparation.

(3–7s): Cooking action with steam/sizzle,
dynamic close-ups, appetizing colors.

(7–10s): Final plated dish, slow camera orbit,
garnish falls in slow-motion, restaurant quality.

Style: Food photography, warm tones, mouth-watering detail.

Common Mistakes vs. Fixes

Mistake	Fix
"A cat video"	"A fluffy orange tabby cat in a sunny kitchen..."
No timecodes	Add `(0–3s):`, `(3–6s):` structure
Too many subjects	Focus on ONE main subject per shot
Conflicting actions	"slow-motion sprint" → pick one speed
Missing camera info	Always specify angle, movement, lens
Vague lighting	"good lighting" → "golden hour, soft shadows"

FAQ

How long should a Sora 2 prompt be?

For simple videos, 1-2 sentences work. For complex, controlled outputs, 100-200 words with timecodes is ideal. The model follows instructions better in shorter clips (5-10 seconds).

Do timecodes actually work?

Yes. Sora 2 reads timecodes and adjusts its generation accordingly. This is one of the most powerful features for maintaining creative control.

Can I use Chinese or Japanese prompts?

Sora 2 works best with English prompts. For other languages, you may get less consistent results. Translate your ideas to English for best output.

What resolution should I specify?

Higher resolution = better quality but more compute cost. Start with 720p for tests, then go 1080p or 4K for final renders. Always specify aspect ratio (16:9, 9:16, 1:1).

My subject keeps changing mid-video. How to fix?

Repeat key physical traits in every timecode section. "The same fluffy white cat" or "the woman in the red coat" helps maintain consistency.

The Bottom Line

Writing great Sora 2 prompts isn't about being creative with words—it's about being specific like a film director.

Three things to remember:

Use timecodes to control shot transitions
Layer your details: scene → subject → action → camera → style
Start simple, then add detail where the model struggles

The best prompts read like storyboards, not poetry.

Ready to generate?

Pro members get instant Sora 2 access, 30+ tested templates, and priority rendering. Start creating today and turn your ideas into cinema-quality videos.