How do you develop a film story, characters, a storyboard and a moving scene with AI without the whole thing looking as if it came off a shelf? Six questions, six answers with worked examples, and the programme for the weekend workshop run by MFVC and vhs Germering on 24 and 25 October 2026.
For the first time, the Munich Film and Video Club and vhs Germering are running a seminar together: “Film with AI: from tool to a voice of your own”, on Saturday 24 and Sunday 25 October 2026, led by Filipa and Michael Campos Viola. This post is the programme preview. It answers the questions people most often bring to us.
In short: AI is not an idea machine. It is a tool that works inside your own research, your own thinking and your own creative stance. Once that sinks in, you get results that look like you rather than like the tool.
Show promptHide prompt
A surreal, humorous Western scene in a sleepy, almost deserted village. Half-wide view down a long dusty main street that runs in strong one-point perspective to the distant horizon. Low, weathered buildings of an old Western town on both sides, wooden facades, porches and a few signs. At the centre stands a surprisingly modern, architecturally elegant pizzeria with large glass surfaces and a warmly lit interior. It looks like a contemporary design object that sits in this old Western town as if it were the most natural thing. Through the windows you can see a modern pizza oven and freshly baked pizzas. Several old horse-drawn carriages and a classic stagecoach stand in front of the pizzeria. A horse waits tied to a wooden post. A single dry tumbleweed rolls across the empty street, driven by the wind. On the extremely wide horizon rise bizarre, slender rock formations and mesas. Above them a huge clear sky with small, stylised clouds. The street and the buildings create a strong graphic vanishing line towards the horizon. The whole world has a premium candy look: slightly heightened pastel colours, warm yellow, orange, turquoise and pink tones, an appetising colour harmony, slightly artificial surfaces but still believable materials and cinematic light. A mix of an elaborately built film set, high-end advertising photography and surreal reality. No modern city, no cars, no crowds. The comic contrast comes solely from the fact that an ultra-modern pizzeria exists in this sleepy Western world as if it were perfectly normal. Cinematic half-wide establishing shot, eye-level camera, strong one-point perspective, 35mm lens feeling, deep depth of field, extremely detailed production design, warm late-afternoon sunlight, long soft shadows, subtle atmospheric haze, premium commercial photography, whimsical visual storytelling, sophisticated candy color palette, photorealistic but slightly heightened reality, cinematic composition, high detail.
Made withOpenArt, GPT 5.6 Sol, Text-zu-Bild
Puzzle 01
Guess the prompt.
This image came from a single prompt, no retouching. Have a good look and work out what you would have had to write to get exactly this. Then open the solution.
- PerspectiveThe street runs to a single vanishing point. No accident: it is in the prompt.
- ColoursPastel, warm, a touch artificial. This look has a name.
- The contradictionOnly one thing does not belong here. That one thing is the joke.
Show solutionHide solution
The prompt, part by part
- Scene
- A surreal, humorous Western scene in a sleepy, almost deserted village. Half-wide view down a long dusty main street that runs in strong one-point perspective to the distant horizon. Low, weathered buildings of an old Western town on both sides, wooden facades, porches and a few signs.
- The contradiction
- At the centre stands a surprisingly modern, architecturally elegant pizzeria with large glass surfaces and a warmly lit interior. It looks like a contemporary design object that sits in this old Western town as if it were the most natural thing. Through the windows you can see a modern pizza oven and freshly baked pizzas.
- Foreground
- Several old horse-drawn carriages and a classic stagecoach stand in front of the pizzeria. A horse waits tied to a wooden post. A single dry tumbleweed rolls across the empty street, driven by the wind.
- Horizon
- On the extremely wide horizon rise bizarre, slender rock formations and mesas. Above them a huge clear sky with small, stylised clouds. The street and the buildings create a strong graphic vanishing line towards the horizon.
- Look
- The whole world has a premium candy look: slightly heightened pastel colours, warm yellow, orange, turquoise and pink tones, an appetising colour harmony, slightly artificial surfaces but still believable materials and cinematic light. A mix of an elaborately built film set, high-end advertising photography and surreal reality.
- Exclusions
- No modern city, no cars, no crowds. The comic contrast comes solely from the fact that an ultra-modern pizzeria exists in this sleepy Western world as if it were perfectly normal.
- Camera
- Cinematic half-wide establishing shot, eye-level camera, strong one-point perspective, 35mm lens feeling, deep depth of field, extremely detailed production design, warm late-afternoon sunlight, long soft shadows, subtle atmospheric haze, premium commercial photography, whimsical visual storytelling, sophisticated candy color palette, photorealistic but slightly heightened reality, cinematic composition, high detail.
The video prompt
Very slow push-in along the one-point perspective of the dusty street towards the glowing pizzeria. The tumbleweed rolls lazily from left to right across the street and out of frame. The horse shifts its weight and flicks its tail once. Warm light inside the pizzeria flickers slightly, heat haze shimmers low over the road. Small stylised clouds drift almost imperceptibly to the right. Nothing else moves. Image-to-video from the provided still. Keep every element of the still exactly as it is: same characters, same faces, same clothes, same objects, same light, same framing. No new objects, no text, no morphing. Slow, cinematic, 5 seconds, seamless loop-friendly, 16:9.
Made withOpenArt, Kling 2.5, Bild-zu-Video, 10 Sekunden, 21:9
Made withOpenArt, GPT 5.6 Sol, Text-zu-Bild
None of that is in the prompt. This is where the real work starts: deciding what stays, what goes, and what you write next time. So, what would you have written?
Further readingFamily Guessing Game AI Fails: Guess the Prompt (Cato Wennison). The same game exists for a games night: 75 AI images, each with its prompt, points for the best guess. English, paperback.
Why do so many AI images and AI texts look alike?
Because a language model or an image generator always delivers the most probable continuation. Without your own material, your own references and clear decisions, everyone ends up with the same middle of the road. The difference comes from what you put in, not from the perfect prompt.
Show promptHide prompt
A large, bright photo gallery with white walls and parquet floor, frontal view of the long main wall. Twenty framed portraits hang on the wall in a perfectly even grid. All twenty portraits show the same smiling young woman in the same pose, the same soft studio light, the same skin tone, as if a single machine had spat them out. Only one picture in the middle breaks the row: a grainy, crooked snapshot of a laughing old woman on a train platform in the rain, with a tear in the paper. A single visitor seen from behind, very small, stands in front of the one different picture and leans forward. A tipped-over paper cup lies on the floor. The gallery wall recedes into depth, the frames get smaller, the room ends in an open door through which grey daylight falls. Cool, clinical gallery look: desaturated whites and greys, smooth surfaces, a hint of turquoise in the light. The one real photo is the only warm, grainy element in the whole picture. No text, no signs, no crowds, no reflections with faces. The joke comes solely from twenty identical pictures and one that is alive. Cinematic wide shot, eye-level, strict frontal symmetry, 35mm lens feeling, deep depth of field, soft even gallery lighting, subtle film grain only on the one odd photograph, premium architectural photography, high detail.
Made withOpenArt, GPT 5.6 Sol, Text-zu-Bild
Puzzle 02
Guess the prompt.
This one came from a single prompt as well. Have a look, have a guess, then open the solution.
- RepetitionCount the pictures. The number is in the prompt, and the AI miscounted.
- The exceptionOne picture is alive. What sets it apart from the other twenty?
- Studio lightCool, smooth, clinical. There is a word for this look.
Show solutionHide solution
The prompt, part by part
- Scene
- A large, bright photo gallery with white walls and parquet floor, frontal view of the long main wall. Twenty framed portraits hang on the wall in a perfectly even grid.
- The contradiction
- All twenty portraits show the same smiling young woman in the same pose, the same soft studio light, the same skin tone, as if a single machine had spat them out. Only one picture in the middle breaks the row: a grainy, crooked snapshot of a laughing old woman on a train platform in the rain, with a tear in the paper.
- Foreground
- A single visitor seen from behind, very small, stands in front of the one different picture and leans forward. A tipped-over paper cup lies on the floor.
- Horizon
- The gallery wall recedes into depth, the frames get smaller, the room ends in an open door through which grey daylight falls.
- Look
- Cool, clinical gallery look: desaturated whites and greys, smooth surfaces, a hint of turquoise in the light. The one real photo is the only warm, grainy element in the whole picture.
- Exclusions
- No text, no signs, no crowds, no reflections with faces. The joke comes solely from twenty identical pictures and one that is alive.
- Camera
- Cinematic wide shot, eye-level, strict frontal symmetry, 35mm lens feeling, deep depth of field, soft even gallery lighting, subtle film grain only on the one odd photograph, premium architectural photography, high detail.
The video prompt
Static camera, then a very slow dolly forward towards the one odd photograph on the wall. The visitor leans in a little closer and tilts his head. The twenty identical portraits remain completely frozen; only the one real photo seems to breathe: a faint flicker of grain on it, as if it were a living image. Spotlights hold steady. Nothing else in the gallery moves. Image-to-video from the provided still. Keep every element of the still exactly as it is: same characters, same faces, same clothes, same objects, same light, same framing. No new objects, no text, no morphing. Slow, cinematic, 5 seconds, seamless loop-friendly, 16:9.
Made withOpenArt, Kling 2.5, Bild-zu-Video, 10 Sekunden, 21:9
Made withOpenArt, GPT 5.6 Sol, Text-zu-Bild
Generative AI has learned from millions of texts and images what “usual” looks like. Ask it something general and you get something general back. That is why Saturday morning starts with precisely this question: how does an AI “think”, why does the answer sound so convincing, and why do you still have to check it? It includes the small exercise that produced the most audible gasps at the club workshop in January.
Example: one question, two answers
You
How many Oscars has Steven Spielberg won? Name the source.
Tool
Language model, for example ChatGPT
Depending on the model and the day, you get three, four or five. Only asking for the source and checking the Academy’s own site gives you the right number. That is the lesson: AI delivers probability, not truth. Checking remains your job.
In the workshop everyone does this at the same time on their own laptop and compares answers.
How do I develop a film story with AI, from idea to logline?
By not asking the AI for a story but handing it yours: the place you know, the observation that will not leave you alone, the conflict that interests you. The AI sorts, asks back, suggests variants. The decision stays with you.
Show promptHide prompt
An empty suburban railway platform shortly after six in the morning, a long line along the platform edge to the signal at the end. Wet concrete, cold blue dawn light, the first sun as a thin strip on the horizon. In the middle of the platform stands a small, perfectly laid breakfast table with a white tablecloth, coffee pot and two cups, as if it were a hotel terrace. An old man around seventy in a grey coat and flat cap sits at it and scatters breadcrumbs for pigeons, as if it were the most natural thing. Twenty pigeons on the platform, one sits on the back of the second, empty chair. On the table an open notebook with a single handwritten sentence, underlined. The writing is too small to read. Tracks run into the dawn, behind them the silhouette of blocks of flats and a water tower, above a clear, cold sky with a single pink strip of cloud. Quiet, cool morning look: blues and grey-blues, one warm spot at the table from a small candle, a light layer of mist over the tracks, photorealistic, calm. No passengers, no train, no legible signs, no big city. The joke comes solely from the laid table on the empty platform. Cinematic wide shot, low eye-level camera, strong one-point perspective along the platform edge, 35mm lens feeling, deep depth of field, cold blue hour light with one warm practical light, subtle mist, photorealistic, high detail.
Made withOpenArt, GPT 5.6 Sol, Text-zu-Bild
Puzzle 03
Guess the prompt.
This one came from a single prompt as well. Have a look, have a guess, then open the solution.
- One sentenceSomething lies open on the table. It is what this chapter is about.
- Early morningBlue, cold, one warm spot. What time does the prompt say?
- PigeonsHow many of them are there? That number is in the prompt too.
Show solutionHide solution
The prompt, part by part
- Scene
- An empty suburban railway platform shortly after six in the morning, a long line along the platform edge to the signal at the end. Wet concrete, cold blue dawn light, the first sun as a thin strip on the horizon.
- The contradiction
- In the middle of the platform stands a small, perfectly laid breakfast table with a white tablecloth, coffee pot and two cups, as if it were a hotel terrace. An old man around seventy in a grey coat and flat cap sits at it and scatters breadcrumbs for pigeons, as if it were the most natural thing.
- Foreground
- Twenty pigeons on the platform, one sits on the back of the second, empty chair. On the table an open notebook with a single handwritten sentence, underlined. The writing is too small to read.
- Horizon
- Tracks run into the dawn, behind them the silhouette of blocks of flats and a water tower, above a clear, cold sky with a single pink strip of cloud.
- Look
- Quiet, cool morning look: blues and grey-blues, one warm spot at the table from a small candle, a light layer of mist over the tracks, photorealistic, calm.
- Exclusions
- No passengers, no train, no legible signs, no big city. The joke comes solely from the laid table on the empty platform.
- Camera
- Cinematic wide shot, low eye-level camera, strong one-point perspective along the platform edge, 35mm lens feeling, deep depth of field, cold blue hour light with one warm practical light, subtle mist, photorealistic, high detail.
The video prompt
Slow lateral tracking shot from left to right along the platform edge, keeping the breakfast table centred. The old man gently scatters crumbs; pigeons bob and peck, two of them flutter up briefly and settle again. The candle flame moves, steam rises from the coffee pot. Thin mist drifts over the tracks, the pink strip of dawn glows a touch brighter. No train arrives. Image-to-video from the provided still. Keep every element of the still exactly as it is: same characters, same faces, same clothes, same objects, same light, same framing. No new objects, no text, no morphing. Slow, cinematic, 5 seconds, seamless loop-friendly, 16:9.
Made withOpenArt, Kling 2.5, Bild-zu-Video, 10 Sekunden, 21:9
Made withOpenArt, GPT 5.6 Sol, Text-zu-Bild
A logline is one sentence: who wants what, what stands in the way, what is at stake. With AI the road there is quicker if you treat it as a sparring partner rather than the author. On Saturday afternoon everyone develops a logline and a short synopsis from an idea of their own. If you already have material, a photo, a newspaper cutting, a few notes, bring it along.
Example: logline from your own observation
You
I observed this in Germering: every morning at six, an elderly man feeds the pigeons at the S-Bahn station, even though a sign says he must not. Ask me five questions that help me find a conflict for a short film. Do not suggest a story yet.
Tool
Language model
The AI asks, for instance: who is he feeding them for? Who minds? What happens on the morning he does not turn up? The logline grows out of the answers, and it is yours, not the machine’s.
How do I develop characters with AI without making them interchangeable?
With contradictions and with details from the real world. A character made only of traits is average. A character with a secret, a habit and a sentence they would never say becomes somebody. AI helps you interrogate the character, not invent the average.
Show promptHide prompt
A small, cluttered tailor's workshop in an old tenement flat, afternoon light through a tall window, bolts of fabric, an old sewing machine, on the wall a pinboard full of notes and photos. At the table sits the old man from chapter 02, grey coat over the chair, flat cap on the table, sewing with utmost concentration a tiny costume for a pigeon that waits patiently on a velvet cushion in front of him. On the pinboard hang sketches of pigeon costumes like fashion designs. A character sheet on the table, half under a pair of scissors: a drawn portrait of the man, beside it index cards with notes, one of them bearing a large red question mark. Thimble, spools of thread, reading glasses. Through the window behind him, out of focus, the platform from chapter 02 with pigeons, as if the workshop were right next to the station. Warm craftsman look: ochre, rust red, dark wood, soft window light, dust in the air, materials with visible texture, photorealistic with a touch of fairy tale. No legible text, no second person, no modern technology except an old lamp. The joke comes solely from a serious old man sewing pigeon costumes. Cinematic medium-wide shot, eye-level camera slightly from the left, 50mm lens feeling, shallow depth of field on the man and the pigeon, warm late-afternoon window light, dust particles, photorealistic, high detail.
Made withOpenArt, GPT 5.6 Sol, Text-zu-Bild
Puzzle 04
Guess the prompt.
This one came from a single prompt as well. Have a look, have a guess, then open the solution.
- SecretWhat exactly is the man doing? And for whom?
- Character sheetPreparatory work lies on the table and hangs on the wall. One piece carries a sign.
- Two livesLook out of the window. Where have you seen that place before?
Show solutionHide solution
The prompt, part by part
- Scene
- A small, cluttered tailor's workshop in an old tenement flat, afternoon light through a tall window, bolts of fabric, an old sewing machine, on the wall a pinboard full of notes and photos.
- The contradiction
- At the table sits the old man from chapter 02, grey coat over the chair, flat cap on the table, sewing with utmost concentration a tiny costume for a pigeon that waits patiently on a velvet cushion in front of him. On the pinboard hang sketches of pigeon costumes like fashion designs.
- Foreground
- A character sheet on the table, half under a pair of scissors: a drawn portrait of the man, beside it index cards with notes, one of them bearing a large red question mark. Thimble, spools of thread, reading glasses.
- Horizon
- Through the window behind him, out of focus, the platform from chapter 02 with pigeons, as if the workshop were right next to the station.
- Look
- Warm craftsman look: ochre, rust red, dark wood, soft window light, dust in the air, materials with visible texture, photorealistic with a touch of fairy tale.
- Exclusions
- No legible text, no second person, no modern technology except an old lamp. The joke comes solely from a serious old man sewing pigeon costumes.
- Camera
- Cinematic medium-wide shot, eye-level camera slightly from the left, 50mm lens feeling, shallow depth of field on the man and the pigeon, warm late-afternoon window light, dust particles, photorealistic, high detail.
The video prompt
Static camera on the tailor, slight handheld breathing. His hands pull the thread through with small, precise movements; he squints and adjusts his glasses once. The pigeon on the velvet cushion turns its head and blinks. Dust drifts through the warm window light. Outside the window, two pigeons on the platform take off in the blur. The desk lamp stays steady. Image-to-video from the provided still. Keep every element of the still exactly as it is: same characters, same faces, same clothes, same objects, same light, same framing. No new objects, no text, no morphing. Slow, cinematic, 5 seconds, seamless loop-friendly, 16:9.
Made withOpenArt, Kling 2.5, Bild-zu-Video, 10 Sekunden, 21:9
Made withOpenArt, GPT 5.6 Sol, Text-zu-Bild
Example: interrogating the character
You
Here is my character: [name, age, job, the conflict from the logline]. Answer as this character in the first person to three questions: what do you hide from your family? What do you do every day without thinking about it? Which sentence would you never say? Keep it short and concrete, no explanations.
Tool
Language model
The answers are suggestions. What fits stays, what does not goes. After ten minutes the character has edges for a story to catch on.
How do I create a storyboard with AI in which the character stays the same?
With a fixed reference image and a fixed description that repeats in every image prompt: age, face, clothes, light, lens. And image-to-image rather than starting from text every time. That keeps the character recognisable across a whole scene.
Show promptHide prompt
A large, slightly worn drawing board seen from above at an angle, on it six storyboard sheets in two rows, black-and-white pencil sketches, between them eraser, ruler, pencil shavings. All six sketches show the same old man in a flat cap on the platform, in six different shots: wide, medium, close-up, over the shoulder, from above, from below. But in each sketch the pigeon beside him wears a different tiny costume: top hat, crown, raincoat, tie, scarf, sunglasses. At the very bottom of the picture, cropped, a real photo of the man as reference, clipped to the top sheet with a paper clip. A hand with a pencil is just starting the seventh sketch, only a few strokes. The upper edge of the picture fades into the blur of the desk, there a coffee cup and a desk lamp casting warm light downwards. Documentary workshop look: paper white, graphite grey, warm lamp light, visible paper texture and pencil grain, photorealistic, the sketches themselves clearly recognisable as pencil drawings. No colour in the sketches, no text, no screens. The joke comes solely from the character staying the same while only the pigeon changes its wardrobe. Top-down three-quarter view, 35mm lens feeling, moderate depth of field with the lower sheets sharp, warm desk-lamp key light from the upper right, soft shadows, premium product photography, high detail.
Made withOpenArt, GPT 5.6 Sol, Text-zu-Bild
Puzzle 05
Guess the prompt.
This one came from a single prompt as well. Have a look, have a guess, then open the solution.
- Six framesSix pictures, six shots. Which ones?
- One characterThe man stays the same. What changes instead?
- PencilSomething is happening at the bottom right. It is in the prompt, word for word.
Show solutionHide solution
The prompt, part by part
- Scene
- A large, slightly worn drawing board seen from above at an angle, on it six storyboard sheets in two rows, black-and-white pencil sketches, between them eraser, ruler, pencil shavings.
- The contradiction
- All six sketches show the same old man in a flat cap on the platform, in six different shots: wide, medium, close-up, over the shoulder, from above, from below. But in each sketch the pigeon beside him wears a different tiny costume: top hat, crown, raincoat, tie, scarf, sunglasses.
- Foreground
- At the very bottom of the picture, cropped, a real photo of the man as reference, clipped to the top sheet with a paper clip. A hand with a pencil is just starting the seventh sketch, only a few strokes.
- Horizon
- The upper edge of the picture fades into the blur of the desk, there a coffee cup and a desk lamp casting warm light downwards.
- Look
- Documentary workshop look: paper white, graphite grey, warm lamp light, visible paper texture and pencil grain, photorealistic, the sketches themselves clearly recognisable as pencil drawings.
- Exclusions
- No colour in the sketches, no text, no screens. The joke comes solely from the character staying the same while only the pigeon changes its wardrobe.
- Camera
- Top-down three-quarter view, 35mm lens feeling, moderate depth of field with the lower sheets sharp, warm desk-lamp key light from the upper right, soft shadows, premium product photography, high detail.
The video prompt
Top-down view, camera slowly drifts down and to the right across the six storyboard sheets. The hand with the pencil continues sketching the seventh frame with small, confident strokes, graphite lines appearing on the paper. A few pencil shavings tremble. The warm lamp light holds. The finished sketches and the reference photo do not change. Image-to-video from the provided still. Keep every element of the still exactly as it is: same characters, same faces, same clothes, same objects, same light, same framing. No new objects, no text, no morphing. Slow, cinematic, 5 seconds, seamless loop-friendly, 16:9.
Made withOpenArt, Kling 2.5, Bild-zu-Video, 10 Sekunden, 21:9
Made withOpenArt, GPT 5.6 Sol, Text-zu-Bild
A storyboard does not have to be pretty. It has to show the shot: wide or close-up, where the camera stands, where the light comes from. Those are precisely the terms that belong in the prompt; then the image generator delivers a readable shot rather than a postcard. On Sunday we turn Saturday’s logline into three to five storyboard frames.
Example: one storyboard frame, described cinematically
You
Storyboard frame, sketch style, black and white. Medium wide shot, camera at eye level, slightly from the left. A man of about seventy, grey coat, flat cap (as in the reference image), stands at six in the morning on the empty platform in Germering, pigeons at his feet, cold morning light from the right. A sign forbidding feeding, blurred, in the background.
Tool
Image generator with reference image, for example OpenArt
A frame that works as an instruction for the shoot. For the next frame only the shot changes; the description of the character stays the same, word for word.
How does a still become a moving scene?
Through image-to-video: the storyboard frame becomes the start frame, and the prompt describes only the movement of camera and character, a few seconds long. It gets even stronger when you upload a clip you shot yourself and change only the look. The movement stays real; the style comes from the tool.
Show promptHide prompt
The same platform as in chapter 02, now in late afternoon with warm backlight. In the foreground a smartphone on a small tabletop tripod, its screen filming exactly the scene behind it. Behind the phone the old man in coat and cap walks along the platform, perfectly ordinary, plastic bag in hand. But on the phone screen you see the same step, the same pose, only as a heroic Western scene: dust, low sun, tumbleweed, the man as a sheriff with a star on his chest. The movement is identical, only the look is different. The tripod, a cable, a power bank, the hand of a filmmaker just touching the record button. Two pigeons watch the man go. Tracks in backlight, blocks of flats as silhouettes, in the phone image instead mesas on the horizon. Two looks in one picture: outside a soft, realistic late afternoon in beige and gold, on the screen the same moment in the candy Western look with heightened pastel colours and haze. No legible text, no other passengers, no montage effects outside the screen. The joke comes solely from the phone showing grey reality as a Western. Cinematic medium shot, camera at hip height behind the phone, 35mm lens feeling, split focus with the phone screen tack sharp and the real scene slightly soft, warm backlight with lens flare, photorealistic, high detail.
Made withOpenArt, GPT 5.6 Sol, Text-zu-Bild
Puzzle 06
Guess the prompt.
This one came from a single prompt as well. Have a look, have a guess, then open the solution.
- TripodThere is a device at the front. What does it show, and what does it leave out?
- Before and afterTwo worlds in one picture. Which one do you already know from puzzle 01?
- The same stepCompare the legs. What is the same, and what is different?
Show solutionHide solution
The prompt, part by part
- Scene
- The same platform as in chapter 02, now in late afternoon with warm backlight. In the foreground a smartphone on a small tabletop tripod, its screen filming exactly the scene behind it.
- The contradiction
- Behind the phone the old man in coat and cap walks along the platform, perfectly ordinary, plastic bag in hand. But on the phone screen you see the same step, the same pose, only as a heroic Western scene: dust, low sun, tumbleweed, the man as a sheriff with a star on his chest. The movement is identical, only the look is different.
- Foreground
- The tripod, a cable, a power bank, the hand of a filmmaker just touching the record button. Two pigeons watch the man go.
- Horizon
- Tracks in backlight, blocks of flats as silhouettes, in the phone image instead mesas on the horizon.
- Look
- Two looks in one picture: outside a soft, realistic late afternoon in beige and gold, on the screen the same moment in the candy Western look with heightened pastel colours and haze.
- Exclusions
- No legible text, no other passengers, no montage effects outside the screen. The joke comes solely from the phone showing grey reality as a Western.
- Camera
- Cinematic medium shot, camera at hip height behind the phone, 35mm lens feeling, split focus with the phone screen tack sharp and the real scene slightly soft, warm backlight with lens flare, photorealistic, high detail.
The video prompt
Static camera behind the phone on the tripod. The old man walks away along the platform with his plastic bag, slow and ordinary. On the phone screen the sheriff version of him makes exactly the same steps at exactly the same pace, dust rising around his boots, the tumbleweed rolling past him. The two pigeons turn to watch him. Lens flare from the low sun pulses gently. Both figures must stay perfectly in sync. Image-to-video from the provided still. Keep every element of the still exactly as it is: same characters, same faces, same clothes, same objects, same light, same framing. No new objects, no text, no morphing. Slow, cinematic, 5 seconds, seamless loop-friendly, 16:9.
Made withOpenArt, Kling 2.5, Bild-zu-Video, 10 Sekunden, 21:9
Made withOpenArt, GPT 5.6 Sol, Text-zu-Bild
Example: your own clip, a new look
You
1. Shoot a short clip on your phone, for instance someone walking along the platform. 2. Upload the clip. 3. Upload a style image, such as the storyboard frame. 4. The tool transfers the look onto the movement. The movement itself stays untouched.
Tool
Video tool with motion control, for example Kling
A few seconds of film that look like the scene from the storyboard, with real movement underneath. In January this was the exercise that made the room loudest.
How do I keep authorship when AI is part of the work?
By keeping every decision with you: which question you ask, which material you feed in, which suggestion you keep, which you throw out. And by checking results, naming sources and labelling what came from the machine. With us, AI images carry a visible label, in this post included.
Show promptHide prompt
The desk from the key visual: evening light through the big window, plants, camera, the black MFVC mug. The whole desk has disappeared under a mountain of printed pages. The pages are almost completely crossed out in red pen, a whole mountain of rejected suggestions, line by line. Only on the top sheet is a single line circled in red. The hand with the red pen rests relaxed beside it, as if this were perfectly normal: a thousand lines gone, one stays. The red pen, almost used up, beside it three more, written dry. A wastepaper basket at the table leg overflows. The pigeon from the other chapters sits on the mountain of paper, a tiny red pen in its beak. Through the window the evening sky over the city, the sun just above the roofs, long shadows across the desk. Warm, cinematic evening look as in the key visual: gold, amber, deep shadows, the red of the strokes as the only cold colour, photorealistic, slight haze in the backlight. No legible words on the pages, only line structure and red strokes, no second person, no screens. The joke comes solely from the mountain of crossed-out pages and the one line that stays. Cinematic medium-wide shot, eye-level camera from across the desk, 35mm lens feeling, shallow depth of field on the circled line, warm golden-hour backlight, long soft shadows, subtle haze, photorealistic, high detail.
Made withOpenArt, GPT 5.6 Sol, Text-zu-Bild
Puzzle 07
Guess the prompt.
This one came from a single prompt as well. Have a look, have a guess, then open the solution.
- Red penCount the pens. What does their state tell you?
- One line staysSomewhere something is not crossed out. Find it.
- Paper mountainWho is sitting on top, and what is in its beak?
Show solutionHide solution
The prompt, part by part
- Scene
- The desk from the key visual: evening light through the big window, plants, camera, the black MFVC mug. The whole desk has disappeared under a mountain of printed pages.
- The contradiction
- The pages are almost completely crossed out in red pen, a whole mountain of rejected suggestions, line by line. Only on the top sheet is a single line circled in red. The hand with the red pen rests relaxed beside it, as if this were perfectly normal: a thousand lines gone, one stays.
- Foreground
- The red pen, almost used up, beside it three more, written dry. A wastepaper basket at the table leg overflows. The pigeon from the other chapters sits on the mountain of paper, a tiny red pen in its beak.
- Horizon
- Through the window the evening sky over the city, the sun just above the roofs, long shadows across the desk.
- Look
- Warm, cinematic evening look as in the key visual: gold, amber, deep shadows, the red of the strokes as the only cold colour, photorealistic, slight haze in the backlight.
- Exclusions
- No legible words on the pages, only line structure and red strokes, no second person, no screens. The joke comes solely from the mountain of crossed-out pages and the one line that stays.
- Camera
- Cinematic medium-wide shot, eye-level camera from across the desk, 35mm lens feeling, shallow depth of field on the circled line, warm golden-hour backlight, long soft shadows, subtle haze, photorealistic, high detail.
The video prompt
Very slow push-in towards the single circled line on the top sheet. The hand sets the red pen down and rests. The pigeon on the paper mountain tilts its head, the tiny red pen in its beak catches the light, then it hops once to the side. Loose sheets stir faintly in a draught. The sun outside sinks almost imperceptibly, the shadows across the desk lengthen. Nothing is written. Image-to-video from the provided still. Keep every element of the still exactly as it is: same characters, same faces, same clothes, same objects, same light, same framing. No new objects, no text, no morphing. Slow, cinematic, 5 seconds, seamless loop-friendly, 16:9.
Made withOpenArt, Kling 2.5, Bild-zu-Video, 10 Sekunden, 21:9
Made withOpenArt, GPT 5.6 Sol, Text-zu-Bild
That also covers the limits: where AI genuinely helps, where it starts making decisions for us, what it costs, and how quickly credits and subscriptions run dry. Responsible use is as much a part of the weekend as experimenting. Sunday ends with everyone showing their own scene, and with a playbook to take home: the working process, the prompt templates and a tool list with prices and privacy notes.
How it came about
The club made the first move itself. In January 2026, eleven members spent a full day at the Stadthalle Germering, laptops on the tables, testing what AI can and cannot do in film. Filipa guided the basics, Michael showed the practice. Until then, many had used AI as little more than a search engine. At five o’clock it had to end, and the questions had not got any fewer. The two-day workshop announced back then is here now, with the vhs as partner and open to everyone.
Real photo, no prompt
The weekend programme
Step 1: Saturday morning: how an AI thinks
A short survey to start, then the basics with Filipa: language models, prompts, why results look alike, and why you have to check. With the exercises from the first block.
Step 2: Saturday afternoon: from your own material to the logline
Michael walks the whole route on one example, then everyone works on their own idea: research, questions to the material, logline, short synopsis, first character.
Step 3: Sunday: from text to image to moving scene
Storyboard frames for your own logline, then image-to-video. Finally limits, costs and authorship. Everyone shows their own scene, and the playbook goes home with you.
Materials & prerequisites
Show promptHide prompt
PromptSame recurring character in all four images: a sympathetic, slightly chaotic man around 38 years old, slim build, messy medium-length dark brown hair sticking out in every direction, short unkempt beard, expressive tired but friendly brown eyes, slightly crooked nose, olive-green corduroy overshirt over a faded burgundy T-shirt, dark worn jeans, mismatched socks, old white sneakers, canvas messenger bag overflowing with papers, pencils and cables. He looks like a brilliant creative person who left home five minutes too late: disorganized, distracted, slightly overwhelmed, but immediately likable. Photorealistic, not caricature, subtle deadpan humor. He stands in the middle of a clean workshop room and proudly holds an open laptop straight out in front of him with both hands like an absurd ceremonial offering. The laptop is clearly his most precious possession. A charging cable hangs ridiculously from it and trails across the floor, tangled around one sneaker. His messenger bag is half open, papers sticking out everywhere. His expression is completely serious, as if presenting an ancient sacred object. The visual joke: despite looking totally unprepared for life, he has brought the one thing that actually matters. No readable text, no logos, no other people. Cinematic medium-wide portrait, eye-level, 35mm lens feeling, warm natural window light, muted sophisticated colors, photorealistic, premium editorial photography, subtle Wes-Anderson-like deadpan composition without imitation of any specific film, high detail, 1:1.
Made withOpenArt, GPT 5.6 Sol, Text-zu-Bild
Your own laptop, essential for taking part
Show promptHide prompt
PromptSame recurring character in all four images: a sympathetic, slightly chaotic man around 38 years old, slim build, messy medium-length dark brown hair sticking out in every direction, short unkempt beard, expressive tired but friendly brown eyes, slightly crooked nose, olive-green corduroy overshirt over a faded burgundy T-shirt, dark worn jeans, mismatched socks, old white sneakers, canvas messenger bag overflowing with papers, pencils and cables. He looks like a brilliant creative person who left home five minutes too late: disorganized, distracted, slightly overwhelmed, but immediately likable. Photorealistic, not caricature, subtle deadpan humor. The man sits at a small table in front of his open laptop. He stares intensely at a completely generic account setup screen with simple empty interface shapes, no readable text and no recognizable brand. He wears his reading glasses incorrectly on top of his head while searching desperately for them with one hand. In his other hand he holds a tiny handwritten password note upside down. Three crumpled attempts lie beside the laptop. His face shows exaggerated concentration, as if creating a free online account were a complicated space mission. Yet the atmosphere remains warm, friendly and reassuring rather than stressful. The visual joke: technology is not the prerequisite, curiosity is. Someone can obviously help this man get started. No readable text, no logos, no recognizable software interface, no other people. Cinematic medium shot, same room, same wardrobe, same character, same visual universe, 35mm lens feeling, warm natural light, photorealistic premium editorial photography, subtle deadpan humor, high detail, 1:1.
Made withOpenArt, GPT 5.6 Sol, Text-zu-Bild
A free account with a language model, for example ChatGPT; we help you set it up
Show promptHide prompt
PromptSame recurring character in all four images: a sympathetic, slightly chaotic man around 38 years old, slim build, messy medium-length dark brown hair sticking out in every direction, short unkempt beard, expressive tired but friendly brown eyes, slightly crooked nose, olive-green corduroy overshirt over a faded burgundy T-shirt, dark worn jeans, mismatched socks, old white sneakers, canvas messenger bag overflowing with papers, pencils and cables. He looks like a brilliant creative person who left home five minutes too late: disorganized, distracted, slightly overwhelmed, but immediately likable. Photorealistic, not caricature, subtle deadpan humor. The man stands at a table and has emptied the entire chaotic contents of his canvas messenger bag onto it. Instead of sophisticated equipment, the table contains wonderfully ordinary personal material: one old printed photograph, a folded newspaper clipping, a tiny notebook with illegible handwriting, a napkin with a rough sketch, a postcard, a dried leaf, several loose handwritten notes and one mysterious photograph he is holding up proudly between two fingers. He looks at this completely unremarkable little photograph with genuine fascination, as if it contained the beginning of the greatest film ever made. Everything around him is chaotic, but his attention is suddenly completely focused on this one small personal object. The visual idea: your starting material does not have to be impressive. It only has to be yours. No readable text, no logos, no other people. Cinematic medium-wide shot, same character and wardrobe, warm directional window light, tactile paper textures, shallow depth of field centered on the photograph and his expression, photorealistic, sophisticated editorial photography, gentle absurd humor, high detail, 1:1.
Made withOpenArt, GPT 5.6 Sol, Text-zu-Bild
An idea, an observation or a piece of material of your own: a photo, a note, a newspaper cutting
Show promptHide prompt
PromptSame recurring character in all four images: a sympathetic, slightly chaotic man around 38 years old, slim build, messy medium-length dark brown hair sticking out in every direction, short unkempt beard, expressive tired but friendly brown eyes, slightly crooked nose, olive-green corduroy overshirt over a faded burgundy T-shirt, dark worn jeans, mismatched socks, old white sneakers, canvas messenger bag overflowing with papers, pencils and cables. He looks like a brilliant creative person who left home five minutes too late: disorganized, distracted, slightly overwhelmed, but immediately likable. Photorealistic, not caricature, subtle deadpan humor. The man sits alone at a workshop table surrounded by intimidating-looking creative equipment: camera, microphone, drawing tablet, cables, notebooks and an open laptop. He clearly has absolutely no idea what any of it does. He holds a computer mouse upside down in one hand and examines it thoughtfully like an unknown archaeological artifact. His other hand is raised slightly in a relaxed shrug. His expression is not embarrassed or frightened, just curious and amused. A pigeon has casually landed on the table beside him and seems to understand the equipment better than he does. His messy bag lies open on the floor, one mismatched sock visible above his sneaker. The visual joke: this is exactly the person the workshop is for. No readable text, no logos, no other people. Cinematic medium-wide portrait, same character, same wardrobe, same room and photographic language as the previous three images, warm natural window light, 35mm lens feeling, photorealistic premium editorial photography, restrained deadpan humor, high detail, 1:1.
Made withOpenArt, GPT 5.6 Sol, Text-zu-Bild
No prior knowledge
Date, venue, registration
Saturday 24 October 2026, 10 am to 4 pm, and Sunday 25 October 2026, 10 am to 2 pm. Venue: vhs in der Harfe 3, Therese-Giehse-Platz 2, 82110 Germering, seminar room 3. Course 262.2445, fee 79 euros, MFVC members 60 euros. Registration is directly with vhs Germering.
Book a place at vhs Germering, course 262.2445
Both days are also in the MFVC calendar, where you can add them straight to your own diary.
Led by Michael Campos Viola, film-maker, photographer and deputy chair of the MFVC, who shows the practice and the sharp end of the technology. Filipa Campos Viola guides participants calmly and clearly through the basics and makes sure nobody gets left behind.
Frequently asked questions about the workshop
Do I need prior knowledge for the “Film with AI” workshop?
No. The workshop starts with the basics. If you already have some experience with ChatGPT or image generators, you are just as welcome as someone opening one for the first time.
What do I need to bring to the workshop?
Your own laptop with internet access, a free account with a language model such as ChatGPT, and ideally an idea or a piece of material of your own to work on.
When and where does the workshop take place?
Saturday 24 October 2026, 10 am to 4 pm, and Sunday 25 October 2026, 10 am to 2 pm, at vhs in der Harfe 3, Therese-Giehse-Platz 2, 82110 Germering, seminar room 3.
What does the workshop cost and how do I register?
79 euros, 60 euros for MFVC members. Registration is via vhs Germering, course 262.2445: vhs-germering.de.
Which AI tools are used in the workshop?
A language model for research, story and characters, an image generator for storyboards and a video tool for the moving scene. Which programmes exactly depends on the state of the art in October 2026; the principles apply to all of them.
Do I take anything home?
Yes: a playbook with the working process, the prompt templates from the weekend and a tool list with prices and privacy notes, plus your own logline, storyboard and short scene.