One formula, six full templates, and eight one-line prompts you can copy, paste, and adapt to your own character.
A prompt is the only creative input that is always available on this site: no upload, no reference clip, and no editing required. That makes wording the highest-leverage change you can make to a generation, and it is also the easiest thing to get wrong by writing too much.
Every example below follows the same five slots - subject, action, setting, light, and camera - and every one is meant to be edited rather than used word for word. Replace the subject with your own upload and keep the structure.
Fill these in order. If you only have time for two of them, keep the action and the camera.
Start with who or what is on screen: the uploaded character, a described person, a mascot, a product, or a creature. When the subject comes from an upload, typing @ and picking it points the prompt at that exact image instead of describing it again.
Give one clear movement the subject performs - walks forward, turns to camera, waves, lifts a product, draws a sword. One primary action per clip reads better than a sequence of three, because a four to eight second shot has room for one idea.
Say where the shot happens and what is behind the subject: a rain-lit street, a bright studio backdrop, a quiet office, a neon arcade. The setting also sets the colour of the shot, so it decides more of the look than most people expect.
Add one lighting instruction - soft key light from the right, warm window light, cold moonlight, neon reflections from below. A single light direction stops the model from inventing flat, front-lit studio lighting on every attempt.
Finish with the framing and the movement: medium shot, close-up, low angle, slow push in, slight handheld drift. Camera wording is the difference between a clip that looks like a test and a clip that looks like a shot.
Each template is written for a four to eight second clip in HD or 720P. Swap the subject, keep the structure, and change one thing per attempt.
The fox mascot in the uploaded image steps out of a poster, turns to camera and waves, bright studio backdrop, soft key light from the right, slow push in from a medium shot to a close-up.
The uploaded presenter speaks to camera in a quiet office, warm window light on the left, gentle head movement, medium close-up with a slight handheld drift, a small nod on the last word.
The uploaded character follows the choreography in the attached dance clip on a rooftop at dusk, matching the timing of the reference, cool blue rim light, wide shot held steady.
The uploaded product sits on a matte pedestal and rotates slowly, seamless dark background, soft top light with a cool rim, slow orbit from a low angle that ends on a three-quarter view.
The knight from the uploaded game art charges forward with a raised sword, dust kicking off the ground, cold moonlight, low angle tracking shot that ends on a close-up of the helmet.
The uploaded portrait speaks the line in the attached audio track, natural head and shoulder movement, quiet indoor light behind the subject, locked medium close-up with no camera movement.
Short prompts are useful when you are testing a direction rather than a specific shot. Paste one, generate, then add detail to whichever part you want to control next.
| Shot | Prompt |
|---|---|
| Slow reveal | The uploaded character turns from the window to face camera, soft daylight from behind, slow dolly in, medium shot. |
| Product spin | The uploaded speaker rotates once on a glass stand, studio backdrop, crisp top light with a soft reflection, locked medium shot. |
| Walk to stop | The uploaded character walks towards camera and stops in a confident stance, wet city street at night with neon reflections, low angle. |
| Idle to action | The uploaded game character shifts from an idle pose into a combat stride, dusty arena, harsh side light, tracking shot that ends wide. |
| Talking head | The uploaded portrait speaks naturally to camera, warm interior light on one side, small head movements, locked close-up. |
| Poster loop | The uploaded poster art gains parallax and drifting light, subtle camera push across the artwork, animated loop, no subject movement. |
| Group formation | Three uploaded characters stand in a line and turn to camera together, cool studio light, slow sideways track, wide shot. |
| Macro detail | The uploaded product tilts to show its side panel, shallow depth of field, warm practical light in the background, macro shot. |
Typing @ inside the prompt opens a picker of everything attached to the generation, and inserting a reference binds it to the words around it. That is how a single prompt can drive one character from a dance clip while another speaks to an audio track.
References are optional, and the generator accepts up to three character references, one motion video of up to 15 seconds, and one audio track of up to 15 seconds. Images can be JPG, PNG, WEBP, HEIC, or HEIF up to 30 MB each; motion can be MP4 or MOV between 2 and 15 seconds and up to 50 MB; audio can be WAV or MP3 between 2 and 15 seconds and up to 15 MB.
The most common cause is a prompt that stops after the subject. When action, setting, light, and camera are missing, the model substitutes its own defaults, and the result reads as generic rather than wrong. Adding one camera sentence and one light sentence usually changes more than rewriting the whole prompt.
The second cause is overloading. A clip that has to walk, turn, wave, and change location in six seconds spends its budget on continuity and delivers none of them cleanly. Keep one action and one camera move, then generate again with a single change if the result is close.
Change one slot per pass and keep everything else identical, so the result tells you which change mattered. If the framing is right but the mood is wrong, edit only the light sentence. If the mood is right and the movement is wrong, edit only the action sentence.
Keep the prompts that worked. A saved prompt plus a saved reference set is the fastest way to produce a consistent series of clips, and it removes the guesswork from the next generation entirely.
Ready to try these in the generator? Start with the how to use Viggle AI guide, or see how prompts behave with a single image in the image to video guide and with no upload at all in the text to video guide.
It is the written direction for one generation: who is on screen, what they do, where the shot happens, how it is lit, and how the camera behaves. Everything except the subject is optional, but each missing part is filled in by the model rather than by you.
One or two sentences is usually enough. Between 20 and 40 words covers subject, action, setting, light, and camera without burying the instruction. Prompts that run to three paragraphs tend to describe several shots at once, and a four to eight second clip can only deliver one.
Typing @ opens a picker of the images, motion clips, and audio tracks you have uploaded, and choosing one inserts it at the cursor. The reference is bound to the words around it, so you can point a motion clip at one character and an audio track at another in the same prompt.
Yes. In a motion transfer prompt the action is already provided by the reference clip, so keep the movement sentence short and spend the words on subject, setting, light, and camera instead. Naming a movement that contradicts the reference clip is the most common reason a transfer looks wrong.
Yes. Text to video works from the prompt alone, and it is the fastest way to test an idea before spending time on reference material. Uploads exist to pin down a specific character, a specific performance, or a specific voice, not to make the prompt work.
Camera instructions get weaker as more of the prompt competes for attention. If a shot ignores a camera move, shorten everything else first and keep a single camera sentence, then generate again. Asking for two camera moves in one clip - a push in and a rotation, for example - usually loses both.
Sign in with Google, drop one of these prompts into the generator, and render a short character clip in up to HD.
Viggle