OurDream AI Prompt Guide

Knowing some basic OurDream AI prompt engineering is key to making the most out of your OurDream AI subscription. It will not only save you a ton of tokens but also open a door with a ton of creative opportunities like generating hentai using the OurDream AI comic generator.

This is a zero-BS OurDream AI prompt guide for beginners. You’ll learn the types of prompts, how to use them, and where to use them.

How Prompting Works in OurDream AI

It doesn’t matter whether you are generating an AI character, generating a video, or an in-chat image — all prompts work the same.

1. You tell OurDream what you want
2. The AI reads your prompt
3. It creates the result based on your instructions
4. If the result is wrong, you change the prompt and try again

If you type a prompt and don’t get what you want, then 9 out of 10 times it is because the prompt is too vague or unnecessarily complicated. OurDream AI specifically says that long complicated prompts are not automatically better.

OurDream’s own help docs make the same point: prompt quality is about clarity, not stuffing in extra keywords.

Prompt quality isn’t about stuffing in more keywords. It’s about giving the right information to the right generator.

And that last part matters, because an image prompt, video prompt, and character prompt need very different information.

The 4 Types of OurDream AI Prompts

OurDream AI has four places where prompts can be used: in the character generator, when generating images, when generating videos, and during roleplay.

Each one behaves slightly differently.

Character Prompts

You use character prompts when creating a character. These prompts describe things that are part of the character, such as their face, body type, eyes, hair, and other physical features.

The important thing to remember is that character prompts should describe permanent features, not temporary ones.

For example, if you add “wearing a hospital uniform” to the character prompt, OurDream may keep trying to generate that outfit later, even when you want your character to wear something different.

Something like “big blue eyes” makes much more sense because that is a feature you normally want the character to keep.

Image Prompts

Image prompts are used to tell OurDream what you want to see in an image.

A simple structure you can use is:

subject → action/pose → environment → clothing/props → lighting → composition → visual style

Put the most important details first. This makes it easier for the image generator to understand what the image should focus on.

Using a structure like this also helps you include enough detail without making the prompt messy or unnecessarily long.

Without a structure:

A woman at the beach at sunset wearing a black dress, cinematic, sitting down, beautiful lighting, realistic photo.

Using the structure:

Woman → sitting on the sand → tropical beach at sunset → black summer dress → warm golden-hour lighting → medium shot, eye level → realistic photography

The second version gives the same basic idea, but each part of the prompt has a clear purpose.

Video Prompts

Video prompts describe what happens over time. An image prompt is like describing a photo, while a video prompt is more like giving directions for a short scene — what the character does, how they move, and what the camera does.

The template I am always using is:

starting state → subject movement → expression/action → camera movement → pacing → ending state

When generating videos on OurDream AI you can actually pick from text-to-video or image-to-video. The second option is recommended, as it’s significantly easier to get what you want.

Chat & Roleplay Prompts

Besides normal chatting and roleplaying, you can also use OOC commands.

OOC stands for “out of character.” These are short instructions you give the AI to control how the chat or roleplay should continue.

You can use them to do things like:

• make replies shorter or longer
• change the tone of the conversation
• move the story forward
• tell the AI to focus on a specific detail
• correct something the AI misunderstood

For example:

(OOC: Keep your replies short and playful.)

The AI then uses that instruction to guide its next replies without making it part of the actual roleplay.

OurDream AI gives you plenty of chat customization options like reply length and writing style, so these commands are rarely needed. That said, they can be helpful sometimes.

Normal Prompts vs JSON Prompts in OurDream AI

JSON prompts are structured prompts used for AI image generation. At first glance, they may look complicated, but they are actually very easy to use. They let you describe an image in much more detail than a regular prompt, which can often lead to a more accurate result.

They are especially useful when you want to copy a pose, style, composition, or even the overall look of another image using your own character.

For example, let’s say you find a great image on Pinterest and want to recreate it in OurDream AI with your own character. A JSON prompt lets you break the image down into smaller details, such as the pose, hand placement, camera angle, clothing, lighting, and background. This gives the AI much clearer instructions about what you want.

I tested this by trying to recreate the image below in OurDream AI. For both tests, I used the exact same settings:

• Vivid 2 image generator
• 9:16 aspect ratio
• 2 generated images
• I picked the best result from each

JSON prompt used:

Create a highly realistic 4:5 vertical social-media photograph closely matching the supplied reference image.


The subject is an adult woman seated sideways on a cream-and-red padded casino chair inside a busy casino gaming floor.


COMPOSITION AND POSE


She sits sideways with her hips angled away from the camera while rotating her upper torso and face back toward the camera.


Her right arm is raised, with the elbow extending outward above shoulder level. Her right hand rests naturally in her hair around the crown/back of her head.


Her left shoulder sits slightly lower. Her left arm falls naturally downward, with her hand resting lightly near her forward thigh.


Her fingers should look relaxed and anatomically correct.


Both legs are bent while seated and extend diagonally toward the lower-left portion of the frame. The lower legs and feet are outside the crop.


Her head is turned toward the camera with a very slight tilt toward the raised arm and a subtly lowered chin.


The posture should feel relaxed and naturally posed rather than rigid.


EXPRESSION


She looks directly toward the camera.


Use:
- relaxed eyes
- subtle cheek lift
- relaxed eyebrows
- small closed-mouth smile
- slight natural asymmetry in the smile
- no visible teeth


FACE


Match the visible facial structure in the reference image closely.


Important characteristics:
- soft oval-to-heart facial outline
- cheek region slightly wider than the jaw
- gradually tapered lower face
- softly rounded, slightly pointed chin
- medium-sized horizontally elongated eyes
- subtly elevated outer eye corners
- medium-thickness groomed eyebrows with gentle arches
- narrow-to-medium straight-looking nose
- softly defined rounded nose tip
- medium-width mouth
- lower lip moderately fuller than upper lip
- soft cupid's bow
- natural left-right facial differences


Preserve the original proportions and feature spacing.


Do not artificially enlarge the eyes, reduce the nose, enlarge the lips, sharpen the jaw, or make the face perfectly symmetrical.


SKIN


Use a warm medium-light to medium tan appearance with golden tones influenced by casino lighting.


Preserve realistic:
- skin texture
- subtle tonal variation
- mild under-eye shading
- slight natural skin sheen
- different illumination across the face, shoulders, arms and legs
- warm light mixed with cool blue reflections


Avoid excessively smooth or plastic-looking skin.


MAKEUP


Polished evening makeup with moderate intensity:
- even complexion
- subtle warm cheek definition
- peachy warm blush
- neutral brown/taupe eye makeup
- dark upper lash definition
- medium-to-long dark lashes
- groomed brown-toned eyebrows
- nude-pink lips
- subtle satin/gloss finish


Keep the makeup natural-looking and believable.


HAIR


Very long, dense blonde hair extending approximately to the waist.


Hair characteristics:
- light golden and beige blonde tones
- darker blonde/light-brown roots
- dimensional highlights and lowlights
- near-center part
- large loose S-shaped waves
- moderate volume at the crown
- several distinct locks falling over both shoulders
- long strands hanging across the front of the torso
- natural irregular strands around the face and crown
- the raised hand slightly lifts and compresses the hair around the crown


Do not turn the hair into one uniform mass. Preserve individual locks, uneven strand placement and realistic gravity.


WARDROBE


She wears an elegant fitted black sleeveless sequined cocktail dress.


Dress characteristics:
- high round neckline
- sleeveless construction
- fitted silhouette
- black stretch backing fabric
- dense reflective black sequins
- blue, cyan, green and silver reflections caused by surrounding casino lights
- realistic fabric tension around the seated waist and hips
- subtle natural folds from sitting
- no visible logos


Accessories:
- delicate necklace
- thin metallic ring on the lowered hand


CASINO ENVIRONMENT


Recreate an authentic indoor casino slot-machine area.


Several large vertically oriented slot machines stand immediately behind the subject.


Important details:
- tall black gaming cabinets
- intense electric-blue illuminated borders
- large colorful displays
- blue aquatic and diamond-inspired graphics
- large jackpot-style numerical displays
- colorful illuminated upper screens
- cyan and blue interface graphics
- glossy black gaming consoles
- touchscreen control panels
- illuminated buttons and payment controls
- additional colorful gaming machines deeper in the room
- empty padded casino chairs
- cream upholstery with red accents
- dark burgundy patterned casino carpet
- geometric faceted ceiling panels in muted warm beige/pink tones
- dark gaps between machine rows
- beverage cans on the gaming console at the right
- small transparent cup containing an orange-colored drink


The environment should remain visually busy and authentic rather than simplified.


FRAMING


Vertical 4:5 composition.


Frame approximately from above the illuminated gaming displays down to below the subject's knees.


The subject sits around the middle of the image.


Her face is located in the upper-middle region.


Large slot machines fill most of the upper background.


The casino chair remains clearly visible behind and underneath her.


Another casino chair is partially visible along the left side.


Part of the gaming console remains visible on the right.


Keep very little empty space.


CAMERA


Make the photograph resemble a genuine image taken by another person using a modern smartphone rear camera.


Camera characteristics:
- photographer approximately 1.5–2 meters away
- camera slightly above the seated subject's eye level
- mild downward viewing angle
- approximately 24–30 mm full-frame-equivalent perspective
- moderately wide smartphone framing
- subtle foreground perspective enlargement
- minimal facial distortion
- relatively sharp background
- no artificial portrait blur


LIGHTING


Use authentic mixed casino lighting.


Combine:
- warm overhead ambient illumination
- bright blue and cyan slot-machine light
- electric-blue LED borders
- warm golden-orange illumination across exposed skin
- cooler blue reflections along the hair, dress and silhouette
- soft frontal ambient facial illumination
- deeper shadows within the hair and beneath the arms
- bright irregular reflections across the sequins
- glossy reflections across gaming-machine surfaces


The white balance should remain naturally mixed rather than perfectly corrected.


IMAGE CHARACTERISTICS


The finished result should resemble an authentic smartphone photograph posted to social media.


Preserve:
- vivid blue and cyan casino illumination
- warm skin rendering
- computational HDR appearance
- local tone mapping
- moderate smartphone sharpening
- slight reduction of very fine texture
- deep environmental shadows
- very bright LED highlights
- mild wide-angle perspective
- subtle social-media compression
- uneven real-world illumination
- natural shoulder-height difference
- slight facial asymmetry
- irregular hair placement
- realistic hand anatomy
- realistic wrist positions
- seated fabric tension
- irregular sequin reflections
- environmental clutter
- screen reflections
- objects partially cut by the edges of the frame


Avoid:
- studio lighting
- heavy background blur
- perfectly symmetrical facial features
- exaggerated facial proportions
- artificial skin smoothing
- generic simplified hair
- changing which arm is raised
- visible teeth
- invented tattoos or piercings
- matte-looking dress fabric
- empty or overly clean casino surroundings
- cinematic studio-style grading


FINAL RESULT


Produce a realistic casino-night social-media portrait matching the supplied reference as closely as possible in composition, pose, facial proportions, expression, hairstyle, black sequined dress, blue-lit casino surroundings, mixed warm-and-cool lighting, camera perspective and natural smartphone-photo imperfections.

Paragraph prompt used:

Create a realistic vertical 4:5 photo of an adult woman sitting sideways on a cream-and-red casino chair in front of bright blue-lit slot machines. She’s wearing a fitted black sleeveless sequined cocktail dress, with very long blonde hair with darker roots and loose waves falling over both shoulders. Her right arm is raised with her hand in her hair, while her left hand rests naturally near her thigh. She turns her upper body and face toward the camera and gives a small relaxed closed-mouth smile. Keep the casino busy and authentic, with tall slot machines, blue LED borders, colorful jackpot screens, patterned carpet, other chairs, drink cans and a small cup on the machine console. The lighting should feel like real casino lighting, with warm golden light on her skin mixed with strong blue and cyan reflections from the machines. Make it look like a genuine smartphone photo taken by another person from slightly above eye level, with a relatively sharp background, mild wide-angle perspective, realistic skin texture, natural facial asymmetry, loose hair strands, small fabric folds, and normal social-media image processing.

At first glance, the image made with the JSON prompt and the one made with the regular prompt look very similar. But when you look closer, the details are different.

The JSON version is much closer to the original image. Her right hand is placed correctly in her hair, and the overall pose is more accurate.

This is where JSON prompts can be useful. A regular prompt may get the general idea right, while a JSON prompt can give you more control over the smaller details.

Does Prompt Structure Actually Matter?

For simple images, not always.

I tested the same idea with prompts that became more and more detailed. Surprisingly, OurDream got the main idea right every time.

That makes sense. If you ask for something simple like “woman drinking coffee in Paris,” there are only a few important things the AI needs to understand.

Prompt structure becomes much more useful when you want specific details to be correct.

1. Basic

woman with blonde hair in sundress

2. Specific

woman with blonde hair in sundress sitting at a cafe in paris and sipping a coffee

3. Structured

woman with blonde hair in sundress sitting at a cafe in paris and sipping a sparkling water, while golden hour, man on motorcycle driving by, muscular black man sitting next to her, menu on table

4. Very Long

woman with blonde hair in a sundress, sitting at an outdoor cafe in Paris, sipping sparkling water from a glass in her left hand, right hand resting on the table, muscular black man sitting beside her, menu lying on the table, man riding a motorcycle past the cafe in the background, warm golden hour sunlight, medium shot, eye-level camera, realistic photography

OurDream is better at understanding simple prompts than I expected. You don’t need to make every prompt extremely detailed.

Add details when you actually care about them. The more specific you are, the more control you have, but you also give the AI more things it can get wrong.

That is a much more useful takeaway than trying to prove that “structured prompts are always better.”

Common OurDream AI Prompting Mistakes

Even a good AI model can give you a bad result if your prompt is confusing. These are some of the mistakes I see most often.

Making Your Prompt Too Vague

The AI cannot know what you have in your head.

If you simply write:

woman at the beach

OurDream has to decide almost everything for you, including the pose, outfit, lighting, camera angle, and what the beach looks like.

If you care about a detail, include it in your prompt. You do not need to describe everything, only the things you actually want control over.

Adding Too Many Instructions

More details do not always mean a better prompt.

If you give OurDream too many instructions at once, it has more things to understand and more things it can get wrong.

Start with the important details first. You can always add more if the first result is missing something.

Putting Temporary Details in a Character Prompt

Character prompts should mainly describe things you want your character to keep.

For example, blue eyes or long blonde hair make sense in a character prompt.

Something like wearing a hospital uniform does not. You may want her wearing something completely different in your next image or roleplay.

Keep temporary things like outfits, locations, and situations out of the permanent character description when possible.

Describing an Image When You Are Generating a Video

A video prompt should explain what happens, not only what the scene looks like.

For example:

blonde woman on a beach, black dress, sunset

This mostly describes an image.

Instead, tell the video generator what should move:

She slowly walks toward the camera while her dress moves in the wind. She then looks over her shoulder as the camera follows her.

Think of an image prompt as describing a photo and a video prompt as giving directions for a short scene.

Changing Too Many Things at Once

If a generation is almost right, do not rewrite the entire prompt.

Let’s say you like the image, but her hand is in the wrong place. Change the instruction about her hand and generate again.

If you change the pose, outfit, camera angle, lighting, and location at the same time, you will not know which change actually helped.

Using the Same Prompt Style for Every Generator

One of the biggest things I learned while testing OurDream is that different image generators can respond differently to the exact same prompt style.

For example, I tested JSON prompts with Vivid 1 and Vivid 2, and both followed them surprisingly well. I could give very detailed instructions and still get an accurate result.

I then tried similar JSON prompts with the default Dreamy generator. The results were much less accurate and the image quality was worse. Dreamy actually worked better for me when I turned the same instructions into a normal paragraph prompt.

So if a prompt works badly, it does not always mean the prompt itself is bad. Try changing how you write it or test it with another generator.

There is no single prompt format that works best with every OurDream model.

My OurDream AI Prompting Workflow

It doesn’t matter if I am creating an image prompt, video prompt, or character prompt. My workflow almost always starts outside of OurDream AI in ChatGPT.

It may sound obvious, but AI is really good at creating prompts for other AI tools.

1. I Explain What I Want to ChatGPT

I normally open a ChatGPT conversation and explain what I want using the voice button.

I prefer talking instead of typing because I naturally give a lot more detail. When I type, I tend to keep things short. When I talk, I explain the idea, what I want it to look like, what is important, and what I don’t want.

You don’t need to make it sound professional. Just explain the idea like you would explain it to another person.

2. I Ask ChatGPT to Turn It Into a Prompt

Once ChatGPT understands what I want, I ask it to turn everything I said into a prompt for OurDream AI.

The prompt will look different depending on what I am creating. An image prompt should describe what you want to see, while a video prompt should focus much more on what happens and how things move.

If I am using a specific OurDream generator, I also tell ChatGPT which one I am using.

3. I Add References When I Have Them

If I am trying to recreate a certain image, pose, style, or scene, I also upload the reference to ChatGPT.

This is especially useful when creating detailed image or video prompts because ChatGPT can help break down things that are easy to miss, like the camera angle, hand position, lighting, background, and composition.

4. I Test the Prompt in OurDream

Then I paste the prompt into OurDream and generate.

I rarely expect the first generation to be perfect. I first look at what it got right and what it got wrong.

5. I Fix Only What Went Wrong

If the result is almost right, I don’t start over.

For example, if everything looks good but the pose is wrong, I go back to ChatGPT and explain exactly what was wrong with the pose.

Then I update that part of the prompt and try again.

This makes it much easier to understand what is actually improving the result.

6. I Save Prompts That Work

When I find a prompt structure that works well, I save it.

After a while, you start building your own templates for different things like images, videos, characters, and certain OurDream generators.

You don’t need to reinvent the prompt every single time.

OurDream AI Prompting FAQ

Do Longer Prompts Give Better Results?

Not always.

A longer prompt gives the AI more information, but it also gives it more things to get wrong.

Use enough detail to explain what you want. Don’t add details just to make the prompt longer.

Should I Use JSON or Normal Prompts?

It depends on the generator.

In my testing, Vivid 1 and Vivid 2 worked very well with detailed JSON prompts. The default Dreamy generator was much less accurate with them and gave me better results with normal paragraph prompts.

So I wouldn’t say JSON is always better. Test both and see what works best with the generator you are using.

Why Is OurDream Not Following My Prompt?

There can be a few reasons.

Your prompt may be too vague, you may be asking for too many things at once, or some of your instructions may conflict with each other.

It can also simply be the generator you are using. If the prompt looks good but the result keeps failing, try another generator or rewrite the same instructions in a different format.

Should I Use the Same Prompt for Every OurDream Generator?

No.

Different generators can respond differently to the same prompt.

A prompt that works perfectly with Vivid 2 may not work as well with Dreamy. That is why I always test the prompt with the generator I actually plan to use.

What Should I Do If My Image Is Almost Perfect?

Change only the part that is wrong.

If the outfit, background, and lighting are good but the hand position is wrong, only change the hand instruction.

If you rewrite everything, you may fix the hand but accidentally ruin everything else.

Should I Put Outfits and Locations in My Character Prompt?

Usually not if they are temporary.

Things like eye color, hair, or other permanent features make sense in a character prompt.

An outfit, location, or situation can change, so those details are usually better added when creating the image or starting the scene.

Do I Need to Know Prompt Engineering to Use OurDream AI?

No.

You can get good results with very simple prompts.

Prompting becomes more useful when you want more control over small details. You can also use ChatGPT to help turn a simple idea into a more detailed prompt.

What Is the Biggest Prompting Mistake Beginners Make?

Trying to make the perfect prompt before generating anything.

Start with what you know you want, generate, look at the result, and then improve it.

It is usually much faster than spending ten minutes trying to write one massive prompt before you even know what the AI will get wrong.