PTM Prepa-Tec Media

Week 2, day 4: animating stills

Making a still picture move.

Students learnt the basics of cinematography first, meaning camera angles, camera movements and shot types, then how to effectively transform their own images into animation: describing subject action, camera motion and background movement as three separate instructions, and judging each result against the still it came from. They also learnt skills like moodboards and LoRAs, small training files that keep a character or a style the same, to work on AI video’s biggest issue, consistency between frames. The shots are shown large, with the frame they started from beside them.

The lab book Week 2, day 4 · skills Before & after Week 2, day 5 · re-rendered shots
Explore
Content generated from AI through an API connection to today's class' Slack channel.

Week 2 · Day 5 · #w2d5_the-oscairs

Before & after

Every film team re-rendered a shot from their trailer, most pushing the same prompt through a stronger video model, then had to defend the difference in front of the class. The question that separates a critic from a fan: when the new clip looks better, what made it better: the model, the prompt, or both?

Re-render a shot Post before & after Show your prompt Defend your verdict

Category 01 · 3 screenings

Same prompt, stronger model

The clean experiment: hold the prompt still, swap the model. Whatever changes between these two clips, the model did it.

The tear in the sunlight

Chapolovers

New model, same prompt
Before
After

The prompt · both renders

Turn this image into a slow, emotional cinematic video. Keep the close-up framing on the man’s face as he looks upward toward the warm sunlight, his expression shifting from pain and exhaustion into hope, gratitude, and quiet relief. A single tear slowly forms in one eye and runs down his cheek, catching the sunlight as it falls. His breathing is soft and shaky, his lips slightly part, and his eyes glisten as if he i…

Critical note

Identical prompt on both renders: the textbook control. Watch the tear: one model animates it, the other only implies it. Any difference you see is the model, not the writing.

“Don’t touch that hanbok!”

The White Hanbok

New model, same prompt
Before
After

The prompt · both renders

Slow push-in. The woman holding the fabric remains frozen, her grip tightening slightly. The seller leans in almost imperceptibly, then stops and says: “Don’t touch that hanbok!, you will regret it!!!. The other woman stays still, watching. The lamp flickers unevenly, shadows crawl across the mirror, and the fabric sways faintly as if moved by a breath. Creeping, unhurried pace.

Critical note

One prompt, two renders, clips literally named Before.mp4 and After.mp4, the tidiest submission in the channel. A dialogue line inside a video prompt is a hard test: check which model actually mouths it.

The man fights the zombie

The Last Station the Cure

New model, same prompt
Before
After Veo 3 fast

Prompt · before

The camera stays still. The character is punching the zombie. The zombie falls down. The pace is quick.

Prompt · after

The camera stays still in place. The man fights the zombie. The zombie falls down. The pace is quick and epic.

Their verdict · verbatim

“We used Veo 3 fast for the after. Its much better and the ai understood the task more. It does not move the camera like tha other. Overall quality is better just for 1 more minute.”

Mila D.

Critical note

Nearly clean: the prompt got two words heavier (“fights”, “epic”) while the model changed, but the team’s verdict is about instruction-following, and “does not move the camera” is a checkable claim. Check it.

Category 02 · 5 screenings

Same tool, sharper prompt

The other lever: hold the tool, rewrite the instruction. These pairs show what prompt craft alone buys, and that longer is not the same as better.

The evidence-camera shot

Nopal in Seol

New prompt, same model
Before
After

Prompt · before

The first animated shot is presented entirely through the lens of a recording camera, later collected by police as evidence. The screen shows timestamp overlays, battery icons, and faint static lines, reinforcing the sense of raw documentation. At first, the environment is empty: a foggy wooded area, faint patrol lights flickering in the distance. The handheld perspective wobbles slightly, as if the device rests on t…

Prompt · after

It has to start with the camera being picked up by the police, then you should see forensics arriving at the scene, and finally the bed being put into an evidence bag. The perspective of the video should always be from the camera. (posted in Spanish: “Tiene que empezar con el movimiento de la cámara siendo recogida por la policía…”)

Critical note

The strongest prompt lesson of the day: a ~280-word cinematic essay lost to three sentences of shot order, written in Spanish. The model needed the sequence of events, not the atmosphere adjectives.

Bridging two stills

Bukhananteam

New prompt, same model
Before
After

Prompt · before

duration: 18 seconds, aspect ratio: 235:1 cinemascope, quality: ultra-photorealistic cinematic produc…

recovered from the file name

Prompt · after

Create a smooth connection between these images that match the lighting, make it natural.

recovered from the file name

Critical note

From spec-sheet to intent. The before prompt orders camera-store settings (duration, aspect ratio, “ultra-photorealistic”); the after says what the shot must DO, which is connect two frames and match the light. Intent won.

The humming girl

Hollowskin Group

New prompt, same model
Before
After

Prompt · before

Make her walk while she is singing a scary song.

recovered from the file name

Prompt · after

Make the girl walk slowly toward the camera while humming a terrifying song with no lyrics, only a disturbing and creepy hum. Her eyes glow yellow. Her movement should feel slow, unsettling, and scary. When she looks directly at the camera, she stops humming for a moment and laughs in a creepy way.

Their verdict · verbatim

“I think its better because the song with the lyrcs was weird, and her eyes were not yellow”

Paula M.

Critical note

A verdict with evidence: the rewrite names exactly what failed before (lyrics where a hum belonged, eyes not yellow) and fixes each by constraint: “no lyrics, only a hum”, “eyes glow yellow”. Evaluation driving iteration.

The title card

Chapolovers

New prompt, same model
Before?
After?

Prompt · before

Animate this title card, make it look cool.

recovered from the file name

Prompt · after

Make it appear on screen.

recovered from the file name

Critical note

“Make it look cool” hands the whole art direction to the model; “make it appear on screen” at least names an action. Neither prompt says WHAT cool looks like. Compare with Aylin’s slot grammar below.

The heist-planning wall

BTS Project

Not labelled
Before?
After?

The prompt · both renders

[CAMERA] Medium-wide shot, slow dolly in, slightly above eye level, organized composition. [ENVIRONMENT] wall covered with photos, venue maps, blueprints, notes, and red string, lit by dim fluorescent and red accent light. [SUBJECT] RM (Asian boy) stands at the center pinning the final photo to the wall, looking like the strategist and leader of the mission. [PACE] Focused, deliberate, strategic, calm before action.…

Critical note

The most disciplined prompt in the channel: four labelled slots (CAMERA, ENVIRONMENT, SUBJECT, PACE) plus a style line. One slot per decision means when a render fails you know which slot to edit. The pair itself arrived unlabelled, so as an experiment it can’t be scored: the prompt is the exhibit.

Category 03 · 2 screenings

Two variables changed: spot the confound

When the prompt AND the model change together, the comparison stops being evidence. One team caught themselves doing it, and that admission is the best science in the channel.

The zombies run

The Last Station the Cure

Both changed
Before
After Veo 3 fast

Prompt · before

The camera is still. No change in angle. Only the zombies run. the pace is extremely quick

Prompt · after

Camera dolly in. The camera moves caotically. The zombies run frantically. The pace is quick and caotic.

Their verdict · verbatim

“The quality is better in the Veo 3 fast, however we did change the camera angle and that impacted in how we liked it better.”

Mila D.

Critical note

They swapped the model AND rewrote the camera move, then flagged it themselves: “however we did change the camera angle and that impacted in how we liked it better.” That sentence, naming your own confound, is the skill this whole page is about.

The training-yard montage

El Tigre de Taekwondo

Both changed
Before
After Seedance 2.0 Mini

Prompt · before

Keep the camera perfectly still in one wide locked-off shot of a gritty outdoor taekwondo training yard, like a professional split-screen clone composite / multiplicity shot. Each version of him is in a different position of the yard doing a different training exercise…

Prompt · after

A determined young man trains for an upcoming taekwondo fight in a gritty hillside urban neighborho…

recovered from the file name

Their verdict · verbatim

“The model made a big big difference”

Guillermo M.

Critical note

“The model made a big big difference”, but the file names show the prompt was rewritten too, from a locked-off multiplicity shot to a free training montage. Maybe the model did it; this pair can’t prove it.

Category 04 · 3 screenings

Screenings missing their lab notes

A before/after with no prompt and no verdict can be enjoyed but not evaluated: the prompt is the method section. These clips are real work; the missing labels are the lesson.

Seoul in salsa verde

Dreemteem

Not labelled
Before
After

No prompt posted. See the critical note.

Critical note

Labelled “after and before”, so the order is attested, but no prompt was posted. Which clip do YOU think is the after, and what would you need to defend that call? Exactly: the prompt.

Untitled pair

Echo

Not labelled
Before?
After?

No prompt posted. See the critical note.

Critical note

Two clips, zero words. Without prompt, order, or verdict there is no experiment here, just footage. Compare how much more the Hollowskin or Nopal in Seol cards let you conclude from the same two-clip format.

The woman in the dark

Independent screening

Not labelled
Before?
After?

The prompt · both renders

A dark cinematic psychological horror scene. A woman stands alone in a shadowy place with low lighting and deep darkness around her. The visual quality must be very high, realistic, detailed, and atmospheric. Her skin should look natural and matte, with no shiny or plastic appearance. Far behind her, a dark shadowy figure appears, clearly separated from her body and standing at a noticeable distance. Inside the shado…

Critical note

The prompt is here and it is thorough: negative constraints (“no shiny or plastic appearance”), spatial layout, even sound design. What’s missing is the labelling: which render is which, and what changed between them.

Off the reel

Screened by link only

Two shots shared as tool links rather than clips. What survives is the most interesting artefact: the one-line ask next to what the tool expanded it into.

He is betraying him

Dreemteem

Watch on Krea ↗

What they typed

the guy in the front turns and punches the suit guy

What the tool ran

A medium side shot of two characters walking. Suddenly, the man in the foreground notices the suited man behind him. He quickly stops, turns 180 degrees with a sharp, fluid motion, takes one aggressive step forward, and throws a powerful right hook straight into the suited man's face. The impact is dramatic, with realistic body recoil, head snapping sideways, suit fabric reacting naturally, and cinematic motion blur. The foreground character keeps a determined, confident expression throughout. Dynamic camera tracking, realistic physics, high-energy action, smooth animation, natural character weight, dramatic lighting, subtle handheld camera shake on impact, ultra-cinematic, action movie choreography, no glitches, no morphing, consistent character appearance.

The hand-to-hand fight

Dreemteem

Watch on Krea ↗

What they typed

(one-line ask, expanded by the tool)

What the tool ran

The two characters suddenly engage in an intense hand-to-hand fight, exchanging fast punches, dodges, and powerful kicks with cinematic choreography, realistic impacts, and dynamic camera movement.