Meme video sits on a knife edge: too clean and the joke evaporates, too rough and the algorithm filters it. The 10 prompts below are built around durable internet-cultural patterns — the ratatouille payoff, the CEO-lifestyle satire, the “30 days of X” timelapse, the chaotic friend-group POV — specific enough that the model lands the comedic beat, loose enough that you can swap the punchline in post.
Two models matter for memes as of June 2026. Google Veo 3.1 (released Jan 13, 2026) added native 9:16 vertical output, 48 kHz synced audio, and the “Ingredients to Video” tool that accepts up to three reference images for consistent faces — ideal for the timelapse and “first time vs now” formats below. OpenAI Sora 2 generates up to 20 seconds at 1080p with synced sound and handles chaotic multi-person scenes well. Both cap a single generation at the 4 / 6 / 8-second marks (Veo) or up to 20s (Sora 2), which suits meme rhythm. See Veo vs Sora video prompts and the image-to-video prompt guide for deeper model setup.
TL;DR
- Keep clips at 4–6 seconds (Veo’s 4s and 6s presets are perfect); one comedic beat per clip.
- Generate the empty top “caption room,” then add the caption in CapCut or Premiere — neither model renders crisp meme text reliably.
- Use Veo 3.1 for micro-expression formats (ratatouille, “trying not to laugh”) and any 30-day / before-after series — its reference-image consistency keeps the same face across cuts.
- Use Sora 2 for chaotic multi-person POV where the mess is the point.
- Lock lens, lighting, and palette across a series; change only the comedic premise.
What a high-quality meme prompt should contain
Five required elements:
- Lens: natural 50mm or first-person 35mm; never anamorphic on a meme
- Light state: soft daylight, fluorescent office, bedroom warm lamp, party-light bedroom; match the meme’s native habitat
- Motion: one obvious comedic beat — zoom, reveal, freeze, timelapse cut; never two beats in one clip
- Palette: slightly warm and saturated; meme video reads “real life,” not “cinematic”
- 4 to 6 second restraint: meme beats land between 4 and 6 seconds; longer kills the rhythm. On Veo 3.1, pick the 4s or 6s duration preset rather than 8s
10 copy-ready prompt templates
1. “Ratatouille moment” reveal
Best for: Food creator, nostalgic-trigger meme
Vertical 9:16, a young person sits at a small kitchen table and takes the first bite of a homemade meal, their eyes widen slightly with genuine surprise as if a childhood memory just hit, soft warm kitchen daylight, plain wall behind, sharp focus on face, room at top for caption, 4 seconds2. “Lifestyle of a CEO” satire
Best for: Office-culture satire, “girl boss” parody
Vertical 9:16, a young person in an oversized blazer sits dramatically at a kitchen table holding a coffee mug with both hands, looks out the window with mock-serious expression as if running a Fortune 500 company, soft morning daylight, plain warm interior, room at top for caption, 4 seconds3. “Trying X for 30 days” timelapse
Best for: Self-improvement meme, fitness, habits
Vertical 9:16, fast cut series of the same young person in the same bedroom corner across 30 days — early days look tired and slumped, later days look brighter and more upright, soft consistent daylight, plain background, sharp focus, room at top for caption, 5 seconds4. “Wait it gets worse” zoom
Best for: Storytime meme, payoff escalation
Vertical 9:16, a young person sits in a small messy bedroom with a slightly defeated expression, slow zoom in on the face as the realization deepens, soft warm interior light, plain background, sharp focus on the eyes, room at top for caption, 5 seconds5. “POV: chaotic friend group”
Best for: Friend-group comedy, party meme
Vertical 9:16 first-person view from one friend at a small living-room couch as three other young friends laugh and gesture wildly around them, soft warm lamp light, slight handheld feel, plain warm interior, sharp focus on the group, room at top for caption, 5 seconds6. “Boss makes you do” satire
Best for: Workplace comedy, corporate-life meme
Vertical 9:16, a young office worker sits at a beige cubicle desk in fluorescent office light with a flat-affect expression as if assigned an absurd task, slow tiny zoom on the face, plain office background, sharp focus, room at top for caption, 4 seconds7. “Explaining X with 1 prop” hook
Best for: Educator-comedy meme, niche-identity reveal
Vertical 9:16, a young creator stands in soft daylight holding up a single ordinary object (a banana, a sock, a mug) with an explanatory expression as if it explains a complicated concept, plain warm wall behind, sharp focus on the face and prop, room at top for caption, 4 seconds8. “Trying not to laugh” challenge
Best for: Reaction meme, duet hook
Vertical 9:16 close-up of a young person facing camera trying to keep a straight face as a smile breaks through in slow motion, soft daylight, plain background, sharp focus on the mouth and eyes, room at top for caption, 4 seconds9. “First time vs now” timelapse
Best for: Skill-progress meme, glow-up content
Vertical 9:16, side-by-side split holds an awkward "first attempt" on the left and a confident "now" version on the right, soft consistent daylight, plain background, slow tiny zoom on the right side, sharp focus, room at top for caption, 5 seconds10. “Side of the internet” reveal
Best for: Niche-community meme, “for you” insider hook
Vertical 9:16, a young person looks at camera with a slightly amused conspiratorial expression and gestures as if "welcome to this corner of the internet", soft warm interior light, plain wall behind, sharp focus on the face, room at top for caption, 3 secondsWhich model for which template
| Template | Best model | Why |
|---|---|---|
| 1 Ratatouille, 8 Trying not to laugh | Veo 3.1 | Cleanest single-face micro-expression shifts |
| 3 30-day, 9 First vs now | Veo 3.1 (Ingredients to Video) | Up to 3 reference images keep the same face across both halves |
| 5 Chaotic friend-group POV | Sora 2 | Handles multi-person motion and overlap when the mess is the point |
| 2 CEO, 6 Boss, 7 One-prop | Either | Single subject, simple staging; pick on cost/access |
| 4 Wait it gets worse, 10 Side of the internet | Veo 3.1 | Slow zoom + held expression render most reliably |
Veo 3.1 vertical (9:16) and 48 kHz synced audio shipped in the Jan 13, 2026 update; Veo generation in the Gemini app sits behind a paid plan (Google AI Pro at $19.99/mo is the cheapest tier with a small daily Veo quota), so budget for a subscription before you batch. Sora 2 outputs up to 20s at 1080p with synced sound — fine for meme clips, just trim to 4–6s on export.
Common mistakes
- Cinematic lens or anamorphic flare — kills the “real life” tone instantly
- Trying to land two jokes in one clip — meme rhythm needs one beat
- Letting AI render the on-screen caption — captions live in CapCut / Premiere, not the model
- Over-styling — meme video reads better when slightly rough
- Past 6 seconds — the joke loses momentum (Sora 2 will happily make 20s; resist it for memes)
How to push results further
- Template 1 (ratatouille) and 8 (trying not to laugh) lean on micro-expression — Veo 3.1 lands the face shifts most reliably
- Template 3 (30-day) and 9 (first vs now) need a consistent palette across both halves — lock
soft consistent daylightandplain backgroundverbatim, and on Veo 3.1 feed the same reference image to both - Templates 2 (CEO) and 6 (boss) read best with slightly absurd props — a too-large mug, a too-small notebook
- Keep trend-specific audio out of the visual prompt; layer the trending sound in post so a copyright takedown never touches your source clip
- Series strategy: lock palette + lens + lighting; swap only the comedic premise
FAQ
Q: How do I keep memes from looking too AI?
A: Slight roughness wins. Use natural 50mm, soft daylight, plain backgrounds, and skip stylization. The closer the visual reads to a phone-shot Reels clip, the better it lands.
Q: Can AI render the on-screen meme caption?
A: Not reliably. Generate the empty top room, then add the caption in CapCut, Premiere or InShot so it stays sharp across platforms.
Q: How fresh do the cultural references need to be?
A: Templates here lean on durable patterns (ratatouille payoff, CEO satire, 30-day timelapse) that survive trend cycles. Swap the punchline in the caption to ride a current trend; the visual stays evergreen.
Q: Best model for chaotic friend-group POV?
A: Sora 2 — it handles overlapping multi-person motion when the chaos is the point. For single-face micro-expression formats (ratatouille, “trying not to laugh”), Veo 3.1 is cleaner.
Q: How do I keep the same face across a 30-day or before-after series?
A: Use Veo 3.1’s “Ingredients to Video,” which accepts up to three reference images and holds character consistency across cuts. Feed the same reference photo to every clip and keep soft consistent daylight and plain background identical.
Q: How long should the meme clip be?
A: 4 to 6 seconds is the sweet spot. 3 seconds for reaction-only hooks, 6 seconds only if the punchline is a slow zoom or reveal. On Veo 3.1 use the 4s or 6s preset; trim Sora 2’s longer output down on export.
Q: What’s the cheapest way to test these?
A: As of June 2026, Veo generation in the Gemini app needs a paid plan — Google AI Pro ($19.99/mo) is the entry tier and includes a small daily Veo quota, enough to prototype a few templates before scaling up. There is no free Veo tier, so plan for a subscription if you want to batch.
Related articles
- Viral Shorts Video Prompts
- Social Ad Hook Video Prompts
- Cinematic AI Video Prompt Examples
- YouTube Thumbnail Prompts
- Viral Shorts sub-hub
- All AI Video Prompts
Tags: #viral-shorts #meme #trending #Video generation #Prompt