Back to blog

Creative workflows

AI Video Trends, November 2026: Seven Tested From One Photo

Seven November AI video trends run on MiniMax H3 Max from one photo each: 80s VHS, action figure, Earth zoom out, pets and more. 19 takes, USD 5.00, one failure.

Nov 10, 2026By Kevin Fan

Video examples here were generated on Cicadas and keep their model, settings and date. Cicadas earns money when you generate with these models. How we test

Test window:
2026-09-27 – 2026-11-12
Clips recorded with these models:
190
Models:
MiniMax H3 Max
Their provider cost:
≈$56.58
Selection:
every clip we keep is listed with its settings, cost and request ID; failed or discarded runs are listed in the test record.
AI Video Trends, November 2026: Seven Tested From One Photo

Short answer: we ran this month's AI video trends from one ordinary photo each on one model, MiniMax H3 Max, and six of the seven worked well enough to show; four of them have a page. The seventh, the bicep heart pose, failed on all four takes. Every published result is below, with the prompt idea, the clip and what differed. Total cost: USD 5.00 for 19 submissions, at 480P and 9:16, at the list price of USD 0.05 per second.

We make Cicadas, so the page links are ours. The trend list came from a public roundup of viral AI video trends, which we checked against search demand on November 10, 2026. The references are the site's own sample photos. What we say a clip shows comes from frame sheets we looked at; sound was measured, not listened to, so we quote what a prompt asked for and do not describe what is heard.

The results at a glance

TrendTakesCostDid it workPage
80s VHS home video3USD 0.75Yes, on take 2 for the living room; the arcade clip has made-up lettering on the signs80s AI video generator
AI action figure box2USD 0.50Yes, with extra items in the box and a figure that outgrows itAI action figure video
Earth zoom out2USD 0.75Yes, with caveats: the scene changes mid-way and the globe's continents are approximateEarth zoom out AI
Pets doing human things2USD 0.50Yes; the laptop screen in the typing clip shows text-like linesDog video, cat video
Car alarm dance4USD 1.00Yes, on take 3 for the dog (take 1 had a badge on the car, take 2 trotted off) and take 1 for the catCar alarm AI dance
Monkey takes your apple2USD 0.50Yes, as an early trend; no page yetNone yet
Bicep heart4USD 1.00No, 0 of 4None built
Total19USD 5.00

The takes are submissions: the 19 are the 19 requests we sent, including the ones we did not publish. For how this model reacts to prompts, see the MiniMax H3 prompt guide; for what a month of samples cost us, see what 255 samples cost us.

80s VHS home video

The trend: a short clip made to look like a 1980s family video on a VHS camcorder, with the set, clothes and hair of the decade. Our prompt idea was one person, one set, one action and a look line of worn tape. We ran three takes; the first living-room take came back clean and modern, and the second added worn-tape wording and was the one we kept.

What differed from the prompt: the picture stays fairly sharp, with no rolling tracking lines, and she wears the photo's own sweater and hair rather than the prompt's pastel sweater and blow-dry. It does have a rounded black frame and a warm cast.

The arcade clip needed one take: he plays, turns, grins and raises a fist under neon. What differed: the cabinet marquees carry made-up lettering although the prompt asked for no text. Both prompts word for word, with the take 1 versus take 2 comparison, are in 1980s photo prompt. Page: 80s AI video generator.

AI action figure box

The trend: a photo of a person turned into a collectible action figure in a toy box. Our prompt idea was a blue blister card with no lettering, or an orange window box on a desk, and a small stiff movement. We ran two takes, one of each, and both are published.

In the blue box, the figure turns its head and lifts a stiff wave while the camera pushes in. What differed: the tray holds two laptops, and the cup carries a small round emblem the prompt did not ask for.

In the orange box, the figure walks out on jointed legs and ends with its hands on its hips. What differed: by the end the figure stands taller than the box it came out of. Page: AI action figure video.

Earth zoom out

The trend: the camera pulls straight up from a person to the street, the city, the coast and the whole planet in one move. Our prompt idea was a single shot with no cuts and no map labels. We ran a 10-second take and a 5-second take.

At 10 seconds, she waves on the grass and the camera rises over the park, the city blocks, the coastline and the clouds, and ends on the globe in space. What differed: the park turns into a different square between two moments instead of one unbroken pull-back, and the continents on the globe are approximate, not accurate geography.

At 5 seconds, a man waves on a rooftop at sunset, and the camera passes streets, a headland and clouds to the globe. What differed: five seconds makes the middle a blur of streets, and the globe's continents are approximate. Page: Earth zoom out AI.

Pets doing human things

The trend: a pet caught doing something a person would do. We ran two ideas from pet photos, one take each: a terrier raiding the fridge at midnight and a tabby typing on a laptop. Both are scenes on our AI dog video generator and AI cat video generator pages.

Midnight fridge: the terrier stands on its hind legs at the open fridge with cheese in its mouth; the light comes on, it turns its head and the cheese drops in the last frames. The tan coat, dark ears and teal collar match the photo.

Typing cat: the tabby sits upright on an office chair with both paws on the keys, looks at the camera for about a second and goes back to typing. What differed: the laptop screen shows rows of text-like lines, where the prompt asked for a plain white glow.

Car alarm dance

The trend: a pet dancing to the beat of a car alarm. Our prompt idea was a pet standing upright on its hind legs in front of a parked car whose indicators flash like an alarm, bouncing from one back paw to the other, waving its front paws in time and ending with them pressed together. The sound line asks for rhythmic car-alarm beeping and light paw taps, with no music, so any sound in the clip is the model's own and not the trend's audio. We ran four takes: three for the terrier and one for the tabby.

The first dog take, on a driveway with a grey car, put a brand-like badge on the car. The second asked for no badge and got a grey car, no spin, and a dog that trots off. For the third we moved the scene to a wet car park at dusk with a red car and ended the dance with the paws together instead of a spin.

Dog, take 3: a wet car park at dusk, a red car side-on with a lit indicator. The terrier stands on its hind legs, bounces, waves one front paw and then the other, and ends with both front paws held together at its chest. The tan coat and teal collar match the photo. What differed: the car shows a door mirror and handles in full, so more than doors and wheels, and the bounce is small.

Cat, take 1: a similar wet car park with a red car. The tabby starts sitting, stands up on its hind legs, waves its front paws up and down, then sits back with both front paws pressed together. What differed: it sits down for the ending instead of staying upright. Page: Car alarm AI dance.

Monkey takes your apple

The trend: a wild monkey takes an apple out of a person's hand. It is early: our trend list had it as new on November 1, and Google autocomplete offered nothing for "monkey eating apple ai" or "monkey steals apple ai" when we checked on November 10, and the keyword tool we use held no rows. With no sign of search demand we built no page, and we plan to check again on November 17. We ran two takes, one with each sample photo, and kept both.

Forest path: she holds a red apple and smiles; a macaque comes in on a branch from the right, reaches over, takes the apple and leaves; she looks at her empty hand and then at the camera with her mouth open. What differed: the monkey is cut by the frame edge for most of its time on screen.

Stone steps: he is about to bite a green apple; a macaque walks up, takes it with both paws, sits next to him and eats it; he looks at his hand, then laughs at the camera. What differed: the apple already shows a bite when the monkey takes it.

The two clips differ in what the monkey does: on the path it snatches the apple from a branch and leaves, and on the steps it walks up and stays to eat it. Neither has a page yet.

Bicep heart: the one that failed

The pose: flex one arm, then curl the other hand over the bicep so that hand and muscle form a heart. It was a fast-rising search term in our keyword data in early November. We ran four takes with two wordings, USD 1.00 in all, and none of them made the pose:

  1. Gym, take 1: he flexes and the other hand rests flat under the bicep. No heart.
  2. Beach, take 1: she makes an ordinary two-hand heart next to her head. Not the pose.
  3. Gym, take 2 (side-on, with a "curved C shape" wording): the other hand points at the bicep with thumb and finger. No heart.
  4. Beach, take 2: the other hand lies flat on the upper arm. No heart.

Two stills from the second takes, at 3.8 seconds:

Frame from the second gym take, 3.8 seconds in: the pose does not form a heart shape

Frame from the second beach take, 3.8 seconds in: the pose does not form a heart shape

Our conclusion: a page that sells this pose would show samples that do not do it, so we did not build one. If you want the trend, film the real pose.

What this says about choosing a trend for AI video

Camera moves and set dressing worked: a pull-back to a globe, a toy box on a shelf, a tape look, a kitchen at night. What did not work was an exact hand pose, where two body parts must meet in one specific shape. The model also fills a scene with things you did not ask for (lettering on signs, a badge on a car, extra laptops in a tray, a car's mirror and door handles), so say what must not appear and look at every surface in the finished clip. Before you build on a trend, run it two or three times from a real photo and watch the whole clip. More trends are on AI video trends.

Questions

Which AI video trends worked from one photo?

Six of the seven we ran: the 80s home video (on the second take for the living room), the action figure box, the Earth zoom out, pets doing human things, the car alarm dance (on the third take for the dog) and a monkey taking an apple. Camera moves and set dressing worked; the bicep heart pose failed on all four takes.

How much did the November trend test cost?

USD 5.00 for 19 submissions on MiniMax H3 Max at 480P, 9:16, at the list price of USD 0.05 per second: eighteen 5-second takes and one 10-second take, 100 seconds in all. Seven of the 19 are not published as samples: the four bicep takes, one clean living-room take and two dog car-alarm takes (one with a badge on the car, one with no spin that trots off).

Why did the bicep heart not work?

In four takes with two wordings, the other hand rested flat under the bicep, made an ordinary two-hand heart, pointed at the bicep, or lay flat on the arm. No take formed a heart from hand and muscle, so we built no page for it.

Can I make the bicep heart myself?

Yes, and it is quicker than prompting it: flex one arm, curl the other hand over the bicep and film it. Then, if you like, use an AI video model for the background or a camera move around your real clip.

Want to try one of them? Start with the 80s AI video generator, the Earth zoom out page or the Car alarm AI dance.

Tools used in this guide