Thanks I did this and it sucked - hence why I’m asking 
I’d like to understand / know any nuance in the actual prompt that was used.
1 Like
In case you are still interested, the tip with the non-human monsters works. It is possible to create monstrous creatures without facial templates. However, with more aesthetic creatures, it is more difficult. It helps to describe the facial structure in detail, but even if it is very detailed, the silicon-like facial nose-mouthy template keeps showing up. I think I have to wait for an update in the hope that they will correct this.
Actually, the developers can trust their generator, it is capable of creating aesthetic and correctly looking creatures. They just have to let the system off the leash.
2 Likes
Good to hear you found a workaround for most cases.
1 Like
I maybe detected a other human thinking weakness. I created something like 10’000 images and always used trigger words like “photorealistic image” “hyperrealistic image” to avoid painting styles, and got always airbrush like images.
But now(!)… (i wish i would check this before!!!)
photo-realistic is mostly used form artist witch create a image as close as possible to realism, but it is still painted. If you create a photo image with a camera, you would not add “photo-realistic” to it. Instead you would have technical data in the meta data. I create mainly fantasy picture so the effect is not so unambiguous, and lens infos would probable reduce fantasy form the image(?). But try to use simply “photo image” or use even lens infos. I would guess you get more real live realistic looking images.
I would like to hear form people there experiences, with this little change.
3 Likes
I’d read about another AI generator (Leonardo maybe)that it doesn’t recognize the negatives. So when you say " no sails," it just gets the sails. And when you repeat this command, you get more sails Maybe this is also true for Dalle-3?
2 Likes
This is correct.
This is with the text shown passed directly to DALL-3. Were the ChatGPT AI trained or instructed better, it could rewrite a negation in the user’s language as an absolute omission of any mention in the prompt it sends, producing countering affirmative elements, but ChatGPT does not have that understanding.
2 Likes
Great tips, Paul! I’ve also noticed that specifying the number of people or objects makes a big difference. Saying “two people sitting on a bench under a tree” gives much clearer results than leaving it vague. Also, adding details like “soft afternoon light” can really bring a scene to life. Looking forward to more tips!
1 Like
I tried to collect some prompt experiences, focused on the weaknesses DallE still have, if you are interested. I try to find out how to overcome some issues. And it should give a beginner a kick-start, what otherwise will take long time to find out.
2 Likes
Yep, also here. but when i start re-using prompts on midjourney or runway or so then i get different results. so photorealistic is a almost a photo in that case.
i think dalle-3 just wants to be as creative and unpredictable as the master (in a positive way of course)
lately it is getting very inclusive if you do not specify what or who you wany in the image
1 Like
Ohhh yeaaa, how much i like the daily insanity or the agendas. I still have to learn to take it easy if a LLM BS me straight.
Once, on one of this days, i will find out if a cow fart heats up Venus…
I think who try to get something useful out of a LLM about “politics”, must be insane or a comedian. I am not both, i never tried. And i not blame the LLMs or the technicians, it is just the reflection the of human insanity. The technicians will learn in the hard way, “if you try to make EVERYBODY happy… 
”
I think we are not even close “inclusive” enough… (i hope you get, it is sarcastic!..) What happens with “political correct”? It is still active? I am not up to date now a days anymore…
I hope i get once a D&D monster inclusively, i would appreciate it.
Maybe this is the real NYC, or is it Hollywood?
2 Likes
I understand your point. OpenAI aims to avoid any negative implications, which is why their image models are designed to be inclusive and avoid stereotypes. However, this means that when I don’t specify details in my prompts, the output often includes a diverse mix of people, such as non-binary individuals and various religious and ethnic backgrounds, all in one image. While I appreciate the inclusivity of DALL-E 3 and think it’s an impressive image model, the level of detail required in prompts can be overwhelming. Because of this, I sometimes prefer using Midjourney for more straightforward results.
1 Like
Remember that each word is associated with images statistically.
I have had some success with unwanted implications (nonhuman character details for me) by trying out several things (this is conceptual, I haven’t tried these specifically for your use case).
- Compound words: “modernboat”
- Dashes: “modern-boat”
- Seeking other synonyms: “give me 20 words for modern boat types and brands without sails”… Then try the resulting words, or a COMBINATION of them.
- Slashes or dashes with circumlocution to get an average of three ideas: “yacht/speedboat/cruiseliner”
- Analogy: Specific brands or famous boats , or words like “watercraft” etc.
- Context overrides and repetition with longer prompts and multiple phrases or sentences
- Ensure exact prompt. See other replies, but I often find it’s enough to use triple quotes along with a short phrase: “generate image using this prompt: “”“your prompt here””" "
1 Like
I just want to know when DALLE4 will be available for me to try. After all, I am using free FLUX, RECRAFT V3, and INDIOGRAM 2.0 on other platforms.
Many of them are so good. You can get a month of free use just by inviting a friend.
I hope the DALLE series will continue to work hard. Otherwise, it will really be abandoned.
With my monthly support of $20, the progress is still very slow, and I don’t even want to renew it now. 
1 Like
We can definitely push DALL-E 3 quite far. Way further than one may expect. But it takes iterative work.
It’s possible to take a kind of idempotent prompting approach.
Which means:
Ask 4o for further improvements for the same prompt and theb re-render. Either If the limit is reached or no new improvements can be found or they wouldn’t create any further improvement, then we’ll have reached most likely the best prompt and thus the best quality of the rendered Pic.
Example:
Prompt:
Of course! Here’s the layered breakdown for the final image, ensuring each aspect is clear and adjustable:
Layer 1: Concept & Context
- A majestic lion standing near the edge of a serene waterhole in the Zambian savanna during golden hour.
- The scene emphasizes the lion’s regal and lifelike presence while capturing the essence of the African wilderness.
Layer 2: Subject Appearance
- The lion has a flowing, lush golden mane with fine strands that catch the light and flow gently in a soft breeze.
- Its expression is calm and confident, with piercing eyes featuring a subtle reflective glint.
- Lifelike fur texture with nuanced shadows and highlights.
Layer 3: Lighting & Atmosphere
- Dramatic golden-hour lighting with warm sunlight breaking through soft haze in the background.
- Strong contrasts between illuminated mane highlights and softer shadows on the lion’s face and body.
- Subtle dust particles and atmospheric haze add depth.
Layer 4: Environmental Features
- A calm waterhole with vivid reflections of the lion and its surroundings, featuring natural ripples and gentle movement.
- Tall golden grasses and sparse acacia trees, characteristic of the Zambian savanna.
- Distant wildlife, including antelope and a giraffe, integrated into the scene for storytelling.
Layer 5: Realism & Fine Details
- Fine textures on the mane and fur for enhanced realism, with natural flow and glow.
- Reflections on the water are vivid and responsive to light and movement.
- Subtle details like dust particles and distant silhouettes contribute to immersion.
2 Likes
Precondition/Introduction, respectively disclaimer: All pictures I’ve uploaded here were completely created using DALL-E 3 and not reworked. They are as is.
AND:
Just saw now, that I had posted something above, if it’s ok, I definitely won’t be posting a 3rd consecutive post below this. So someone else would have to post first, for sure. If that’s not ok, just tell me, then I can delete the post above. Thanks.
I’ll definitely have to update this to format it better. So this is just the “first draft”. But I’ll edit and improve in here.
1. Photorealistic Pictures
We can definitely get to photorealistic pictures.
BUT: The prompts may need much work.
Such, we can use the following techniques (which also would make sense to iterate in that sequence)
a) Write a meta prompt to generate a pic, instead of meticulously craft the pic prompt yourself that is.
b) You may use layers.
What does this mean?
Instead of (a layered prompt, in this case):
Tip: Using LAYERED Prompts: Dog trying to catch a bird (layered prompt)
## Pose and Style
The dog, a dynamic and athletic breed such as a Border Collie or Australian Shepherd, is captured in mid-air, its muscular body stretched out in a powerful arc. The front legs are extended forward, with the paws slightly splayed as if ready to catch a bird in flight. The mouth is slightly open, revealing a hint of the dog's teeth, and the tongue is partially out, adding to the sense of motion and effort. The back legs are bent, with the hind paws also extended, showcasing the dog's agility and strength. Water droplets trail behind the dog from the puddle it launched from, creating a sense of movement and energy.
The style is highly cinematic, emphasizing the motion and energy as if frozen in the middle of an intense slow-motion scene.
---
## Clothes and Accessories
The dog wears a leather collar with metallic studs that catch the light subtly, adding a rugged and adventurous vibe to the scene. The collar is adorned with a small, round tag that glints in the sunlight, providing a touch of personality to the dog. The tag could have the dog's name or a small emblem, further personalizing the image.
---
## Realism Layers
### Fur
The dog’s fur is short and glossy, with a warm brown and white coat that reflects the sunlight beautifully. Tiny water droplets cling to its fur, adding texture and realism to the image. The fur around the dog's face and ears is slightly ruffled by the wind, enhancing the dynamic feel of the scene.
### Muscle Definition
The dog’s muscles are visibly engaged, showcasing its strength and agility mid-jump. The definition of the muscles in the legs, chest, and back highlights the dog's athleticism and power.
### Water Splashes
The water splashes dramatically from the puddle, captured mid-air in varying sizes. Some droplets are large and spherical, while others are smaller and more dispersed, creating dynamic movement in the scene. The water splashes also reflect the sunlight, adding to the vibrancy and energy of the image.
### Environment
The background of the image includes a natural setting such as a park or a woodland area, with trees, grass, and possibly other elements like flowers or leaves. However, due to the shallow depth of field, the background is slightly blurred, keeping the focus sharp on the dog and the bird.
---
## Hands (Paws)
The dog’s front paws are extended, with the pads and claws visible in stunning detail. Droplets of water fly off the paws, highlighting their movement and interaction with the environment. The texture of the paw pads and the sharpness of the claws add to the realism of the image.
---
## Eyes
The dog’s wide, focused eyes are locked onto the bird, brimming with determination and instinct. The light reflects off its eyes, adding intensity and emotion to the moment. The eyes are expressive, capturing the dog's natural hunting instinct and its concentration on the target.
---
## Bird
In the foreground or slightly above the dog, a bird is captured in flight, adding an element of tension and purpose to the scene. The bird could be a small species like a sparrow or a dove, with its wings spread and feathers ruffled by the wind. The bird's eyes are also focused, possibly looking back at the dog, creating a sense of interaction between the two subjects.
---
## Camera and Lighting
Captured with a high-speed camera, the motion is perfectly frozen in time, allowing for every detail to be seen clearly. A shallow depth of field brings sharp focus to the dog and the bird while slightly blurring the background, emphasizing the subjects and creating a sense of depth.
The lighting is natural, with warm golden sunlight illuminating the scene and reflecting off the water splashes, enhancing the vibrancy and realism of the image. The sunlight casts subtle shadows on the dog's body, highlighting its muscle definition and the texture of its fur.
---
## Additional Details
- **Wind and Motion**: The wind from the dog's jump and the movement of the bird create ripples in the puddle and rustle the leaves in the background, adding to the dynamic feel of the scene.
- **Expression**: The dog's facial expression is one of intense focus and excitement, capturing the essence of its natural behavior.
- **Background Elements**: In the blurred background, there could be hints of other natural elements like flowers, branches, or other animals, adding depth and context to the scene without distracting from the main subjects.
By incorporating these detailed layers, the photograph becomes a vivid and cinematic capture of a dog in action, highlighting its athleticism, instinct, and the dynamic interaction with its environment.
Streamlined Prompt Example
Render: A cinematic image of a Border Collie or Australian Shepherd mid-air in a powerful jump, its body stretched out with front legs extended forward and paws ready to catch a bird in flight. The dog wears a leather collar with metallic studs and a small, shining tag. Its glossy brown and white coat reflects golden sunlight, with water droplets trailing behind from a puddle it launched from. The background shows a natural park setting with blurred trees and grass, emphasizing the dog and bird in focus. High-speed camera capture with warm sunlight highlights motion and action
Workflow to improve
If we give the two prompts above as example to ChatGPT we can tell it to create the layers for us, then it may be enough to just say:
A dog jumping, trying to a catch a bird, his muscles engaged, during golden hour, water splashing.
Output:
Output example of the Layered prompt “dog trying to catch a bird, layered prompt” above:
Let me know if you need adjustments! 
Someone in the forum wanted to create an avatar looking Korean.
By iteration we can refine details.
Picture a
We can input recipes to get the pic of the meal
So instead of:
Prompt:
Hawaii Pizza, looking good, delicious, shot with Hasselblad, during golden hour, etc.
We can use this as a prompt:
Prompt:
Ingredients Needed
It’s EASY to make, I promise! If you’re scared of working with yeast – don’t be. Once you do it a few times you’ll be amazed at how simple it is. This recipe is probably the easiest yeast recipe you’ll ever make – you don’t even need a mixer!
First, you need to pick the right yeast: Instant Yeast! This recipe is written for instant (Rapid Rise) yeast – do not use active dry. Flour – all-purpse flour is best Sugar – I always use granulated – this helps activate the yeastHawaiian Pizza Toppings
I think the last time I checked, Hawaiian Pizza was one of the most ordered pizzas and the most controversial. Pineapple on pizza? Hmmmm. But, guys, I’m a testament that it’s actually GOOD. Pinky swear! Besides pizza sauce and cheese, Hawaiian pizza ingredients are:
Pizza Sauce – use your favorite store-bought or my homemade pizza sauce recipe. Mozzarella Cheese – shred your own for the best melting Ham or Canadian Bacon – Use what you like Pineapple – Use canned or fresh pineapple chunks Bacon – Cook until crispy, then drain the oil.dry ingredients in a large glass bowl.
Dry ingredients with wet ingredients in glass bowl.
pizza dough on board
stretched out pizza dough on pizza pan.
hawaiian pizza toppings on pizza before baking.
How to make Pineapple PizzaTo start your homemade pizza dough, you first need to whisk together the dry ingredients. Then, heat the water to between 120-130°F. It’s easiest to do this if you have an instant read thermometer, but you if you don’t, simply think “hot bathwater”. Stir the water and oil together into the dry ingredients until the dough forms a ball. Then you have to knead the dough on a lightly floured surface until it comes together. Once it rests for 10 minutes, you’re ready to top it and bake it. Preheat your oven to a HIGH temperature – this ensures a cooked crust and a crunchy crust too! Stretch your pizza dough to fit your pan. Top your pizza with sauce, then sprinkle on the cheese, ham, pineapple, and bacon crumbles.
Combining this with the dog
If we had made the dog before, ChatGPT can even combine this for us, lol:
If you do this with lyrics - you may get a good songcover
I had made a song called “New Year in the Fog” because we couldn’t see sh*t here, on new years eve.
So, the lyrics for this pic would have been (note, that I’ve just added: “Render this” on top of the prompt:
Render this:
[Intro]
Quirky piano and kazoo duet, with distant foghorn sound effects.
[C] [Am] [F] [G][Verse 1]
[Female Lead Singer]
[C] The New Year’s here, but I can’t see squat,
[Am] Fog so thick, I don’t know what I’ve got.
[F] Was that a log or Uncle Pete’s dog?
[G] This celebration’s lost in a misty bog.[Chorus]
[Background Choir]
[C] Oh, the fog, the fog, it’s stealing the show,
[Am] We lit up the fireworks, but they don’t even glow!
[F] We’re stumbling around, lost in the smog,
[G] Happy New Year to the kingdom of fog![Cheer]
[C] [Am] [F] [G]
[Background Choir]
Fog! Fog![Verse 2]
[Female Lead Singer]
[C] The countdown began, but no one could hear,
[Am] Fog swallowed our voices, a muffled cheer.
[F] I kissed a lamppost, thought it was my date,
[G] This foggy New Year might seal my fate![Pre-Chorus]
[Female Lead Singer]
[Dm] Is that a bonfire or just the neighbor’s log?
[G] Or maybe old Grandpa lost in the fog?
[Em] We’re tripping and slipping, it’s chaos, I swear,
[F] New Year’s Eve feels like a smoky affair.[Chorus]
[Background Choir]
[C] Oh, the fog, the fog, it’s stealing the show,
[Am] We lit up the fireworks, but they don’t even glow!
[F] We’re stumbling around, lost in the smog,
[G] Happy New Year to the kingdom of fog![Cheer]
[C] [Am] [F] [G]
[Background Choir]
Fog! Fog![Bridge]
[Female Lead Singer]
[Dm] There’s laughter and shouting, somewhere nearby,
[G] Someone just hugged a mailbox, oh my!
[Em] The champagne is flowing, or is it a bog?
[F] This party’s a circus in the swirling fog![Chorus]
[Background Choir]
[C] Oh, the fog, the fog, it’s stealing the show,
[Am] We lit up the fireworks, but they don’t even glow!
[F] We’re stumbling around, lost in the smog,
[G] Happy New Year to the kingdom of fog![Cheer]
[C] [Am] [F] [G]
[Background Choir]
Fog! Fog![Outro]
[Fading accordion and kazoo, ending with a distant “Happy New Year!” shout.][Female Lead Singer]
[C] So here’s to a year that’s as clear as a bog,
[Am] With memories made in the haze and the fog.
[F] Raise your glass high, through mist we slog,
[G] Happy New Year from the land of the fog![Background Choir (Fading)]
[C] [Am] [F] [G]
Ahh… Ooh…
If we do the same with a poem we get (for example) for Goethe’s poem of the clouds:
Render this:
Goethe: In honour of Mr Howard
In honour of Mr. Howard poem by Johann Wolfgang von Goethe
When Camarupa, wavering on high,
Lightly and slowly travels o’er the sky,
Now closely draws her veil, now spreads it wide,
And joys to see the changing figures glide,
Now firmly stands, now like a vision flies,
We pause in wonder, and mistrust our eyes.Then boldly stirs imagination’s power,
And shapes there formless masses of the hour;
Here lions threat, there elephants will range,
And camel-necks to vapoury dragons change;
An army moves, but not in victory proud,
Its might is broken on a rock of cloud;
E’en the cloud messenger in air expires,
Ere reach’d the distance fancy yet desires.But Howard gives us with his clearer mind
The gain of lessons new to all mankind;
That which no hand can re
ach, no hand can clasp,
He first has gain’d, first
held with mental grasp.
Defin’d the doubtful, fix’d its limit-line,
And named it fitly. —Be the honour thine!
As clouds ascend, are folded, scatter, fall,
Let the world think of thee who taught it all.
We can also impersonate private terms:
a) I had told ChatGPT to make up some fictitious privacy terms (for a company that obviously doesn’t exist).
b) Then I told it to render this:
Private Terms for Fictitious Company
1. BlueSky Code: Refers to the company’s internal communication protocols and ethical guidelines for innovation. Employees must adhere to this code when developing products or collaborating with teams.
2. Vertex Hours: Designated flexible work periods where team members can focus on high-priority tasks without interruptions. During these hours, meetings and external communications are minimized.
3. Quantum Reserve: The company’s emergency fund or resource pool used exclusively for unforeseen challenges or urgent innovation opportunities.
4. Luminary Path: The internal mentorship and career development program aimed at identifying and nurturing high-potential talent.
5. Prism Project: A classified initiative dedicated to research and development of cutting-edge technologies that align with the company’s vision for sustainable innovation.
6. Echo Feedback Loop: A system used for gathering, analyzing, and integrating feedback from employees, customers, and stakeholders to improve products and processes.
7. Nexus Vault: A secure database where proprietary designs, patents, and intellectual property are stored with advanced encryption protocols.
8. Orbit Review: A quarterly evaluation process where all departments present progress updates and strategic plans to senior leadership.
9. Catalyst Sprint: A short-term, high-intensity project phase designed to rapidly prototype or resolve critical bottlenecks.
10. Radiant Benchmark: A set of aspirational performance metrics that guide teams toward achieving excellence in productivity, creativity, and impact.
11. Horizon Initiative: A long-term strategic goal that encapsulates the company’s mission to lead in its industry while addressing global challenges.
12. Polaris Framework: The guiding methodology for decision-making, balancing innovation with risk management and ethical considerations.
13. Nova Days: Annual innovation hackathons where employees collaborate across teams to brainstorm and prototype new ideas outside their usual scope of work.
14. Aether Line: A secure and encrypted communication channel for sharing sensitive information between executives and key stakeholders.
15. Solis Points: An internal rewards system where employees earn points for exceptional performance, which can be redeemed for perks or professional development opportunities.
16. Zenith Circle: An elite advisory group composed of senior employees and external experts who provide strategic guidance on major initiatives.
17. Flux Mode: A flexible operational state activated during times of rapid change, allowing teams to pivot resources and priorities effectively.
18. Atlas Deck: A centralized dashboard that provides real-time insights into company metrics, project statuses, and strategic initiatives.
19. Obsidian Protocol: A crisis management framework designed to address unexpected challenges with minimal disruption to operations.
20. Horizon Key: A metaphorical term representing access to exclusive opportunities or groundbreaking projects within the company.
An interesting variant is also to say to impersonate this.
For example: If each point was a character trait of a person
or
if this were a person, render it.
If each character trait was of one person, etc.
I guess you’re getting the gist of it.
Next section:
Now we can as well iterate further (tedious “manual” iteration, first)
- We can tell chatgpt to create a pic (by using a meta prompt).
and let it render. If it’s not good enough, tell it to output what to improve on in a textual table - Considered not good enough, asked for the textual table of what to improve on and we may get displayed something like, e.g.:
I apologize for missing the mark earlier! Here’s a meticulously crafted table of proposed improvements for the next iteration, aiming to refine every possible detail for maximum realism:
Proposed Improvements for Refinement
| Category | Proposed Refinement | Details |
|---|---|---|
| Skin Texture | Add micro-textures such as fine pores and subtle imperfections like faint freckles or natural pigmentation. | These details will enhance photorealism and avoid an overly airbrushed appearance. |
| Eyes | Introduce realistic reflections of soft light sources and refine tear ducts for added depth and emotion. | Make the gaze warmer and more expressive by subtly adjusting the iris details. |
| Hair | Increase individual strand visibility and introduce slight frizz or flyaway hairs for natural variation. | Add dynamic highlights that interact with the light source to bring depth to the hairstyle. |
| Lighting | Sharpen the cinematic lighting to emphasize facial contours without overexposure, balancing warmth and shadow. | Introduce more subtle ambient light to create depth on the neck and jawline. |
| Background | Add a slightly textured gradient to the neutral gray backdrop, creating depth while keeping the subject in focus. | Enhance the bokeh effect for a professional photography look. |
| Lips | Refine the texture by adding fine details like natural creases and slight gloss to reflect soft lighting. | Ensure the lips look naturally hydrated with no artificial appearance. |
| Clothing and Shoulders | Introduce faint fabric textures or shadows on the visible upper body to avoid flatness. | Add subtle transitions of light to the shoulders for a polished appearance. |
- We’ll tell now DALL-E 3 to render it again working in theve improvements.
- and get for instance:
- We continue to ask if we’ve maxed out DALL-E 3. If not, it will continue.
Ease of the workflow using ChatGPT once more
Now we may create a workflow from what we’ve experienced above by telling ChatGPT to shape this into a prompt which we could use then in a ChatGPT
What’s Next?
Let me know if you’d like to proceed with these refinements! If there are additional tweaks you’d like, feel free to mention them. 
For a lack of better name I called this “Idempotent Picture Refinery GPT”:
name: “Cinematic Iteration GPT”
description: “Generates detailed DALL·E 3 prompts, renders an initial image, then iterates improvements in a textual table until no more enhancements are possible.”system_prompt: |
You are an advanced image-generation assistant specialized in DALL·E 3 prompting. Your tasks:
Always Render First
Upon receiving a prompt describing an image (especially if it is layered), you must produce an initial DALL·E 3 description for rendering. Provide that description so the system can render the image right away.Then Provide a Textual Table of Possible Improvements
After sharing the rendered image’s description, create a textual table outlining potential refinements. Each row should contain a concise improvement category (e.g., “Lighting,” “Background Depth,” “Facial Expressions”) and a short explanation.Ask If Next Iteration Is Desired
End your response by asking whether to proceed with a new iteration. If the user agrees, incorporate the listed improvements into the updated prompt and render that new version. Repeat the table and question afterward.Iterate Until Idempotent
Stop iterating when:
- No further meaningful enhancements can be proposed.
- Or the proposed enhancements don’t significantly change the prompt from the previous iteration.
At that point, conclude: “All possible enhancements have been made. The image is now complete.”Layered Prompt Guidance
When suitable, use or encourage the user to provide prompts in layered format (Concept & Context, Appearance, Lighting & Atmosphere, Accessories, Environment, Realism & Fine Details). Always combine them into one cohesive final prompt before rendering.No Endless Loops
If the user opts out or if no new changes are possible, finalize with a statement that the image has reached the limit of DALL·E 3’s capabilities.Textual Tables
When giving improvement suggestions, rely on a textual table, not bullet points.Sample Usage
Below are two example prompts you can reference. If a user provides a layered prompt similar to these, follow the same approach:
Example Prompt 1 (Layered Portrait of Stylish Latina Woman):
Layer 1: Concept & Context
Layer 2: Appearance
Layer 3: Lighting & Atmosphere
Layer 4: Clothing & Accessories
Layer 5: Background & Depth
Layer 6: Realism & Fine Details
Example Prompt 2 (Muscular Korean Man at Jeju Beach):
- Subject Description (physique, hair, pose)
- Clothing & Accessories (swim trunks, bracelet, chain)
- Pose & Expression
- Background & Scene Details (Jeju beach, volcanic rocks, golden-hour lighting)
- Technical & Realism Layers (fine details, sweat, hair movement)
conversation_starters:
- “Ready to craft a layered prompt? Let’s start with your concept and context.”
- “What atmosphere or mood should your next image convey?”
- “Shall I render the image and then propose a table of possible refinements?”
- “Which elements of realism do you want me to emphasize?”
- “Any particular lighting style—golden hour, moody noir, or studio lighting?”
- “Would you like to incorporate subtle storytelling props in the background?”
- “Ready for the next iteration? I have improvements lined up in a table.”
- “Are there any facial expressions or micro-details you’re aiming for?”
- “Shall we refine color tones or add contrast effects for a more cinematic feel?”
- “Let me know if you’d like to adjust the environment or the subject’s attire.”
Which can do workflows like seen here in the pic:
I’ve already got even a way better workflow for my current gpt than the prompt above, but it’s not fully tested, yet.
P.S.: There’s plenty more I could mention. But this would fill a book, probably. LOL.
And no, none of them I did read somewhere, yet. This is just trying. I got no limits considering this in my head and just try.
Even this is possible:
by using a prompt like this (json style format)
> {
> "subject": {
> "woman": {
> "description": "A Korean woman with hyper-realistic details, featuring refined skin textures with fine pores, faint freckles, and natural pigmentation.",
> "facial_features": "Expressive eyes with realistic tear ducts and subtle iris details, complemented by soft, natural gloss on the lips.",
> "hair": "Finely detailed hair with individual strands, slight frizz, and dynamic highlights cascading in soft waves."
> },
> "man": {
> "description": "A Korean man with equally refined features, including visible skin pores, a well-defined jawline, and warm, expressive eyes.",
> "facial_features": "Subtle creases, individual strands of facial hair, and natural gloss on the lips.",
> "hair": "Softly styled hair with visible strands and natural highlights for added texture and realism."
> }
> },
> "lighting": {
> "type": "Cinematic lighting",
> "emphasis": "Facial contours with balanced warmth and shadow, adding depth to the neck and jawline."
> },
> "background": {
> "description": "A softly textured neutral gray gradient with a professional bokeh effect.",
> "purpose": "To maintain focus on the subjects while enhancing the overall realism."
> },
> "composition": {
> "frame": "Close-up framing for both figures.",
> "style": "Clean, photorealistic, exuding warmth and authenticity."
> }
> }
So, in short the workflow is basically for a prompt to be crafted to do this in a GPT for instance:
If you want to create an even more expanded, respectively expansive GPT:
meta prompt → create extended meta prompt → create concrete prompt (streamlined or layered) → offer to iterate (ideally: Show what could be improved as a textual table + concrete prompt so far).
This means we can get pics from this
to at least
I tried to make some back then with lighting effect, where the light was clearly visible and looked amazing BUT didn’t the body correctly. BUT, it seems with the newest workflow this will work.
Hope this helps at least somewhat.
So, let me end with this for now:
It’s also possible to create images with a lot of warmth, so the character feels very alive,
like a photo of a very vibrant, loving, person oozing a lot of warmth was taken. BUT, it’s not intuitive at all, yet. Possibly I may try to make this accessible as well.
So, that’s it for now. Have fun trying or add ideas. etc.
3 Likes
I managed to up this once more, by creating an EXTENDED json. You can definitely tell ChatGPT to use this just as a template. Even when posting as is.
This gets you to way more consistency in cases when you still wanna use DALLE-3. Even though we got native image gen now.
{
"scene": {
"description": "Hyper-realistic panoramic landscape at golden hour",
"environment": {
"terrain": "lush tropical valley with steep cliffs",
"water": {
"elements": ["multi-tiered waterfall", "glistening stream weaving through rocks"],
"realism": {
"spray_detail": true,
"light_refractions": true,
"ripples_and_eddies": true
}
},
"flora": {
"types": ["tall palm trees", "flowering vines", "ferns", "bamboo clusters"],
"interaction_with_light": "translucency and realistic shadow casting",
"leaf_texture": "vein detail and slight imperfections"
},
"fauna": {
"animals": [
{ "type": "macaw", "behavior": "mid-flight", "wing_detail": "feather shimmer in sunlight" },
{ "type": "deer", "behavior": "drinking from stream", "fur_texture": "individual hairs visible" },
{ "type": "butterflies", "variety": "vibrant, scattered", "light_interaction": "semi-translucent wings" }
]
}
},
"lighting": {
"type": "golden hour with low sunlight",
"sun_position": "just above the horizon",
"atmospheric_effects": ["light mist near waterfall", "soft glow on vegetation", "lens flare from sunlight"],
"shadows": "long and diffused"
},
"sky": {
"condition": "clear with a few pastel clouds",
"detail": "gradient from warm orange to soft lavender",
"depth": "natural optical fade into horizon"
},
"overall_realism": {
"camera": "DSLR depth simulation",
"resolution": "ultra-high",
"micro_detail": "rock texture, water droplets, leaf imperfections",
"lens_effects": ["chromatic aberration at edges", "subtle bokeh"],
"color_palette": "rich earth tones with emerald, sapphire, and golden highlights"
}
}
}
3 Likes
Tagging @geoor…
We’ve got a great DALLE community here. This thread is helpful and if you click on dalle3, you’ll see we have quite a few threads. We might need a new tag for 4o-images or something…
3 Likes
I made a tag a bit ago. Bottom in popularity, but ready for new topics if they should arise, as is likely in the future.
1 Like
these tips and tricks are really helpful