What you want
The scene, the subject and what it’s doing.
How to write prompts that get better video results
A prompt tells DomoAI what to make, how it should look and what to leave out. Learn the four parts of a strong prompt, how to adapt it for each tool, and how to fix weak prompts with the free checker below.
A prompt does three jobs. Leave one out and the AI fills the gap with a guess.
The scene, the subject and what it’s doing.
Style, lighting, camera and colors.
Blur, distortion, text, extra characters and anything else you don’t want.
Click any example to copy it.
Who or what, where, the action and how close the camera is.
The art direction: medium, palette and mood.
How it’s “shot”: lens, light, grade and camera movement.
What to leave out. It’s the easiest layer to forget.
The same four layers apply everywhere, but each tool needs a different focus.
Keep the motion, change the look. Describe the footage as it is, then the new style.
[scene summary] + [style] + [technical look] + [negatives]“Person walking through a sunny park, anime style, soft pastel colors, clean line art, gentle lighting, cinematic framing. No text, no extra characters, no distortion.”
The image already sets the look. Use the prompt for motion, camera and mood.
[action] + [camera move] + [mood] + [negatives]“Smooth looped animation of the character waving and smiling, soft anime style, bright daylight, subtle camera sway. No extra characters, no text.”
Define appearance, clothing, setting and atmosphere so the character stays consistent.
[who] + [look & outfit] + [setting] + [lighting] + [negatives]“Friendly virtual host, short brown hair, casual hoodie, standing in a minimal studio with soft blue lighting, clean modern design. No logos, no text.”
Give every uploaded reference a job, and say what it must not change.
[scene] + [Image 1 = …] + [Image 2 = … ONLY] + [Video 1 = …]“Image 1 is the literal first frame. Image 2 controls the character’s face, hairstyle and wardrobe ONLY. Do not recreate its framing, composition, or lighting.”
Paste a prompt to see which layers it’s missing, plus vague words and clashing instructions. Then add the missing parts in one click.
A quick keyword-based check that runs in your browser. It can’t predict exact results, but it catches the gaps that most often cause weak output.
Start with the core idea, test it, then add one layer at a time. When a result changes, you’ll know which words caused it.
girl playing piano, anime style
girl playing piano, anime style, warm sunset lighting
girl playing piano, anime style, warm sunset lighting, soft pastel colors, gentle camera movement
girl playing piano, anime style, warm sunset lighting, soft pastel colors, gentle camera movement, no extra characters, no logos
In Video to Video, describe what’s actually in the footage. Asking for a car chase over a clip of someone at a desk won’t work.
Better: “person sitting at a desk, anime style, futuristic neon office, shallow depth of field”
For a series of videos, reuse the same style line in every prompt and change only the scene.
Style lock: “anime style, pastel colors, soft lighting, gentle film grain”
1–3 short sentences: scene, then style, then technical and negatives. Long prompts tend to contradict themselves.
One subject, one action and one camera move per clip. Make longer stories from several short clips.
Keep prompts that worked in a notes file. Your own tested library beats starting from scratch.
Start with 34 proven prompts →For exact moves or a consistent face, upload a reference instead of writing a longer prompt.
Motion & Style Reference →Replace the HIGHLIGHTED words, then copy.
A CHARACTER playing INSTRUMENT on a small stage, camera gently moving. Soft anime style, glowing stage lights, pastel color palette, subtle film grain. No extra characters, no text, no distortion.
Close-up of PRODUCT on a rotating platform, dark studio background. Photoreal 3D style, dramatic side lighting, high contrast, reflections on the surface. No hands, no text, no background clutter.
Person speaking to camera, medium shot. Clean modern style, bright daylight from a window, soft depth of field, warm color grading. No text, no on-screen graphics, no flickering.
CHARACTER walking through a forest, turned into fantasy anime style. Magical lighting, glowing particles in the air, rich saturated colors, gentle camera shake. No modern objects, no text, no distortion.
Presenter standing in front of abstract shapes, minimal background. Flat, clean illustration style, corporate color palette (BLUE, WHITE, ACCENT), smooth camera moves. No clutter, no extra characters, no logos.
“Make it cool.”
“Cinematic 3D style, teal-and-orange color grading, dramatic lighting.”
“Dark moody scene with bright cheerful colors.”
“Dark moody scene, deep blue shadows, single warm lamp.”
“Epic car chase in the rain” (footage: someone at a desk)
“Person at a desk, anime style, futuristic neon office.”
“Girl dancing in a park, anime style.”
“…anime style. No text, no extra characters, no distortion.”
Use 3–5 s clips at standard resolution.
Style, then lighting, then camera, one at a time.
Put versions next to each other to see which words made the difference.
Rephrase what almost worked instead of starting over.
On Standard and above, run test batches without spending credits.
Add it to your prompt library with notes on the settings.
The text instruction you give DomoAI. A good prompt says what you want (scene, subject, action), how it should look (style, lighting, camera, colors) and what to avoid (blur, distortion, text, extra characters).
Usually 1–3 short sentences: the scene first, then the style, then technical details and negatives. Longer prompts often contradict themselves and give less coherent results.
Yes. Ending your prompt with exclusions such as “no text, no extra characters, no distortion, no flickering” reduces common problems. It’s the layer people most often forget.
Describe what’s really in your footage, then the new style. Use this pattern: scene summary + style + technical look + negatives. Don’t ask for actions that aren’t in the clip.
Use a style lock: one fixed style line, such as “anime style, pastel colors, soft lighting, gentle film grain”, reused in every prompt. Change only the scene. For consistent faces, add reference images too.
The prompt is probably too long, contradictory (for example, dark and bright at once) or asks for too many actions in one clip. Simplify to one subject, one action and one camera move, then add details one at a time.
It’s safer to describe the look instead, for example “soft watercolor, muted greens, hand-drawn line art”. Descriptions travel better between models and avoid rights issues with commercial work.
Use the templates on this page, or browse 34 copy-paste prompts by category on our Best Prompts & Examples page.
Write your prompt, check it here, then bring it to DomoAI. Start with free credits.