01
Who
Who is on screen?
Give each person an age, a look and one detail the camera can see. A name alone tells the model nothing.
- “a woman in her sixties with silver braids”
- “a scruffy golden retriever”
- “a tiny dragon”
New accounts get one free video· make it count
How to promptlike a film director.
An AI model only sees your words. Every detail you leave out, from the place and the people to the camera and the light, becomes a guess. Watch a plain idea turn into a prompt that leaves nothing to chance, then build your own.
- 8
- ingredients in a prompt that leaves nothing to guess
- 5s
- your free video: room for one clear action
- 1,000
- characters fit in AI Studio's prompt box
Your video prompt64/1,000
Two friends in a coffee shop having an interesting conversation.
Left for the AI to guess:who they arethe place and timetheir clothes and drinkswhat happensthe camerathe lightingthe stylethe sound
0/8
Prompt maker
Your turn. Build it step by step.
One question at a time. Tap an idea or write your own, and the prompt is written as you go, with a coach that flags anything likely to waste your free video.
Start from an example:
An age, a look and one detail the camera can see.
Ideas
The recipe
Eight ingredients. Zero guesswork.
Weak prompts say who and what. Strong ones answer eight questions, in roughly this order. You won't need all eight every time, but each one you skip is a choice the model makes for you.
Who in Where & when, wearing Clothes & props, Action. Camera. Light. Look. Sound.
02
Where & when
Where are we, and when?
Name the place, the time of day and the weather, then add one small detail that makes it feel real.
- “a cozy café at night”
- “rain streaking the window”
- “a foggy pier at dawn”
03
Clothes & props
What are they wearing and holding?
Colors, fabrics and objects stop a scene looking generic, and they keep a character recognizable from clip to clip.
- “a mustard knit sweater”
- “a latte with leaf art”
- “a battered leather satchel”
04
Action
What happens in these few seconds?
One clear action with a beat: something starts and something changes. A clip is only a few seconds long.
- “leans in and whispers”
- “bursts out laughing”
- “sneezes a puff of fire and panics”
05
Camera
Where is the camera, and does it move?
Say how close the shot is, the angle and the movement. Without them the model picks for you.
- “medium two-shot at eye level”
- “slow push-in”
- “low angle”
06
Light
What lights the scene?
Name the light source and its color. Light sets the mood faster than any adjective.
- “warm tungsten lamps”
- “golden hour sun from the left”
- “blue neon reflections”
07
Look
What should it look like?
Pick one style and a color palette, then choose the matching style setting, which is added to your prompt.
- “35mm film, soft grain”
- “teal and amber colors”
- “hand-drawn anime”
08
Sound
What do we hear?
Only for models that make sound: Seedance 2.0 and newer in AI Studio. Your free Seedance Lite clip is silent, so skip this one for it.
- “rain on the glass”
- “an espresso machine hissing”
- “soft jazz”
Real prompts, real videos
Short prompts work when every word works.
Two clips we made for the brief of AITube's “Monsters Need Therapy Too” Prompt Battle, from exactly these words on Seedance 1 Pro with the Cinematic style. Each covers who, where, the action, the camera and the light. Each still left one thing to chance.
Seedance 1 Pro · 5 sThe tiny dragon
A tiny dragon on a therapist's couch sneezes a puff of fire and panics, cozy office at night, slow push-in, warm lamplight
WhoWhere & whenActionCameraLightStyle: Cinematic
- What it got
- The sneeze, the panic, the slow push-in and the warm lamp against a night-time city.
- What it guessed
- “Therapist's couch” became a living-room sofa. A visible detail, like a framed diploma on the wall, would have made it an office.
Seedance 1 Pro · 5 sThe nervous yeti
A huge furry monster squeezed into a tiny therapist's office nervously sips tea from a tiny cup while the therapist takes notes, warm light
WhoWhere & whenActionClothes & propsLightStyle: Cinematic
- What it got
- The office, the tiny teacup, the therapist writing notes and the monster's nervous sip.
- What it guessed
- “Squeezed into” didn't land: the room looks a normal size. Size needs something to compare against, like “its head brushing the ceiling”.
Camera
Tell the camera where to stand.
How close, from where, and how it moves: these are the words that make a clip feel directed instead of random. Copy the phrases in quotes straight into your prompt.
How close
Shot size decides what the viewer pays attention to. Wide shows the world; close shows the feeling.
“wide shot”Shows the whole scene: where we are and who's in it.“medium shot, waist up”Body language with a little of the place behind.“close-up of her face”Emotion. The face fills the frame.“extreme close-up of her eyes”One detail, big: eyes, hands, a drop of water.From where
The camera's height changes how we feel about the subject. Leave it out and you usually get eye level.
“at eye level”Neutral and natural, like being in the room.“low-angle shot looking up”Makes the subject look big, brave or powerful.“high-angle shot looking down”Makes the subject look small or alone.“top-down overhead shot”Straight down: food, maps, crowds, patterns.How it moves
Pick one move per clip and say how fast. Slow moves read as deliberate; fast ones as action.
“static camera”Only the scene moves. Calm and observant.“slow push-in”Pulls you closer and builds focus or tension.“slow pull-back reveal”Starts close and ends wide to show where we are.“slow pan from left to right”Sweeps across a scene from one spot.“crane shot rising up”Lifts up and away for a grand ending.“handheld camera”A little shake that feels real and urgent.More words that work: tracking shot (moves alongside the subject), orbit (circles around it), drone shot (flies over the scene), shallow depth of field (blurred background), macro (tiny details up close).
AITube Music is coming soon
Same recipe, for songs.
“A happy song” leaves everything to chance, just like “two friends in a coffee shop”. Name the genre, the mood, the tempo, the instruments, the voice, the shape of the song and what it's about.
Your song prompt
A song about a summer road trip with friends.
Left for the AI to guess:the genrethe moodthe tempothe instrumentsthe voicethe structurewhat it's abouthow it should sound
0/8
A real one
The prompt behind AITube's own soundtrack
Product film underscore0:00 / 0:42
Instrumental only: no vocals, no vocal chops, no choir, no spoken words. A polished, modern, optimistic underscore for a 3-minute software product tour video with a narrator speaking on top the whole time. Tempo about 104 BPM, D major. Warm analog-style synth pads, soft plucked synth arpeggios (kalimba/marimba-like), muted electric piano chords, deep soft sub bass, light clean percussion (soft kick, shaker, rim clicks, no cymbal crashes). Keep the vocal midrange uncluttered: no lead melody, no busy riffs, no solos, even dynamics, so a female voiceover stays clear above it. Structure: a gentle 8-second intro of pads and plucks; then the beat enters softly; a steady, confident groove with subtle evolving layers every 16 bars (no dramatic drops or builds); in the final 15 seconds resolve to a satisfying final chord with a natural decay and a clean ending. Mood: clever, friendly, premium, calm confidence, like an Apple or Stripe product film. Duration about 3 minutes.
The first 42 seconds of the track this exact prompt made with Google's Lyria model. It plays under AITube's product films. Notice how much of it says what to leave out, so the narrator stays clear.
Label your lyrics
Write lyrics in sections tagged [Verse], [Chorus] and [Bridge]. Many AI music tools read the tags and shape the song around them.
Short lines, one hook
Keep lines short enough to sing in one breath, and repeat one memorable line in every chorus.
Describe, don't name-drop
Describe the sound instead of naming an artist. Many tools refuse artist names, and a description gets you closer anyway.
One genre, two at most
'Lo-fi hip hop with jazz piano' works. Five genres in one prompt blur into nothing in particular.
Write your own in the song prompt maker, or make background music today with AudioForge in AI Studio.
Fixes
Six habits that waste a good idea.
Almost every disappointing first video comes down to one of these. Each takes seconds to fix.
Swap vague words for things you can see
Instead ofAn amazing, epic video of a beautiful city
TryA rain-soaked neon street in Tokyo at night, steam rising from a ramen stall
A camera can't film an adjective. “Amazing” and “epic” leave every choice to the model.
One shot per clip
Instead ofShe wakes up, drives to work, gives a speech and celebrates with friends
TryShe steps onto a stage and the crowd rises to applaud her
A clip has room for one action. Build longer stories scene by scene.
Make a multi-scene ShortSay what you want, not what you don't
Instead ofA street with no people and no cars
TryAn empty street at dawn, completely deserted
Naming people and cars puts them in the picture. AITube has no negative-prompt box, so describe the result.
What is a negative prompt?Add on-screen words afterwards
Instead ofA shop sign that says “Grand Opening: 50% off”
TryA shop window with a big blank red banner
AI video still misspells most text. Add titles and captions in AITube's AI Video Editor instead.
Invent characters, don't borrow them
Instead ofA famous singer dancing on stage
TryA silver-haired pop star in a sequined jacket dancing under spotlights
Real people and famous characters break AITube's rules on deepfakes and rights, and many models refuse them.
Pick one style, then match the setting
Instead ofPhotorealistic anime claymation of a fox
TryA fox in a snowy forest, hand-drawn anime look
Mixed styles pull in different directions. The style you pick is added to your prompt, so make them agree.
Prompts to try
Start from these. Then make them yours.
Six complete prompts. Use one as it is, or keep the structure and swap in your own subject, place and colors. “Use this prompt” opens it ready to generate, and on your free video it picks the matching style too.
Product adStyle: Cinematic
Product shot of a white running sneaker turning slowly on a wet black stone pedestal, water droplets bouncing off the sole, a thin ring of red light glowing behind it. Slow orbit at shoe level, shallow depth of field. Dark studio, a single soft light from above, glossy high-contrast commercial look.
FoodStyle: Realistic
Close-up of a chef's hands pouring glossy warm chocolate sauce over a stack of fluffy pancakes topped with fresh raspberries, steam rising, on a rustic wooden table by a sunny kitchen window. Slow push-in, shallow depth of field, warm morning light, rich food-commercial colors.
NatureStyle: Realistic
Aerial drone shot gliding low over a turquoise river that winds through a pine forest at sunrise, mist lifting off the water, a lone red canoe drifting downstream. The camera glides forward and tilts down toward the canoe. Soft golden light from the left, long shadows, crisp natural colors.
Sci-fiStyle: Cinematic
A lone astronaut in a scuffed white suit walks across a field of glowing blue crystals on an alien planet, two huge moons low on the horizon, dust drifting in the air. Wide shot, the camera tracks alongside at a low angle. Cold rim light from the moons, purple and teal colors, epic sci-fi film look.
CartoonStyle: 3D Cartoon
A chubby orange cat in a tiny chef's hat flips a pancake in a cozy cartoon kitchen; the pancake lands on its head and the cat freezes, wide-eyed. Medium shot at eye level, static camera. Soft warm light, bright saturated colors, playful family-film animation.
AnimeStyle: Anime
A girl with a long red scarf stands on a city rooftop at dusk as the wind blows cherry blossom petals past her, city lights flickering on below. Low angle, slow push-in as she turns to look over her shoulder. Pink and violet sunset sky, soft glow, hand-drawn anime look.
FAQ
Questions about prompting
How long should an AI video prompt be?
Long enough to cover who, where, what they wear, what happens, the camera, the light and the look: usually 40 to 120 words. Length alone doesn't help, so make every word describe something the camera can see or hear. AI Studio's form takes up to 1,000 characters, about 150 words.
Will AITube rewrite my prompt?
Mostly not. For your free video, the model gets your words as you wrote them, with the style you picked added at the end. In AI Studio's chat, the assistant uses a detailed description word for word and only fills in what's missing, such as framing or lighting. In Director, a shot is rewritten only when you click Improve prompt.
What does my free video include?
One 5-second clip on Seedance Lite at 480p, in the aspect ratio you choose. It has no sound, so spend your words on the picture. Each account gets one; Creator and Pro plans come with monthly credits instead.
Why doesn't my video show the words I asked for?
AI video models still misspell most text on signs, labels and captions. Ask for a blank sign or banner, then add the words in AITube's AI Video Editor.
How do I keep the same character across several videos?
Write one description of the character, covering face, hair, clothes and colors, and paste it word for word into every prompt. Models don't remember earlier clips, so small wording changes give you a different-looking person. In AI Studio's chat you can also attach an image to use as the starting frame.
Can I use a celebrity or a famous character in my prompt?
No. AITube's rules prohibit deepfakes and content that uses someone's likeness without consent, and you need the rights to anything you publish. Many models refuse these prompts anyway. Describe an original character instead.
Does a longer prompt cost more credits?
No. Credits depend on the model, the length and the quality you choose, never on how long your prompt is.
Do the same tips work for AI music?
Yes. A song prompt needs the same specifics: genre and era, mood, tempo, instruments, vocals, structure and what the song is about. AITube Music is coming soon; until then, AudioForge in AI Studio makes background music from a description.
Going deeper: prompt engineering, text-to-video and negative prompts in the glossary.
Your idea is ready. Now direct it.
Who, where, what they wear, what happens, the camera, the light and the look. Write those down and press generate.