Prompt discussion - your prompt - and a recurring problem.

 

Sometimes things don't go the way you imagined, and the AI ​​keeps making an interpretation that's wrong for you, which gives you the feeling that the AI ​​is confused. The question then arises... is that really the case? Well, no, it isn't, because the AI ​​is only executing a command, and yes, there can be a difference in the AI ​​model for set settings, but generally you can overcome this by choosing a model that has no trouble with the set prompt by default, but then what's wrong?

On this page we will parse a prompt .- and there is the task: to give the prompt straight lines on all sides , dilemma the ai produces curves that do not improve in a geometric result 

but this never works, and why is that?

Before we dive in, an example.

You give a clear command via the prompt. But, it doesn't do what you want. You want straight lines across all sides of the image > straight lines...

But what could be improved is that here you see an image of a prompt that gives a geometric view... BUT what you see is a round image, and that's not exactly what you had in mind...

 Before we look at the prompt that generated this: 

Style Fusion

 Art Nouveau — geometry

 

Color Palette

- Deep Black #000000 — mystery, border

- Muted Gold #D4AF37 — exaltation, symbolism

- Teal #008080 — synthesis, digital clarity

- Midnight Blue #191970 — night, depth, rhythm

- deep red #7B1E1E -Ceremony,




a woman on the right, positioned on the right side of the image, facing inward (toward the left). She has long flowing hair and wears a decorative headband and an ornate dress with embroidery. The background is empty. Add, geometric gold lines framing all four borders of the image — top, bottom, left, and right. the woman on the right and empty space on the left side of the image. lines all around the frame in the image 


If you look closely, this isn't exactly a top prompt. We see grammatical errors, misinterpretations, and the whole thing can produce a variety of images. First, we'll look for typos and edit to prevent the AI ​​from interpreting things unnecessarily.

an edited prompt produces this:

A woman on the right, positioned on the right side of the image, facing inward (toward the left). She has long, loose hair and is wearing a decorative headband and an ornate, embroidered dress. The background is blank. Add geometric gold lines that frame all four edges of the image: top, bottom, left, and right. The woman on the right and the empty space on the left side of the image. Lines surround the frame within the image.

A woman on the right, positioned on the right side of the image, facing inward (toward the left). She has long, loose hair and is wearing a decorative headband and an ornate, embroidered dress. The background is blank. Add geometric gold lines that frame all four edges of the image: top, bottom, left, and right. The woman on the right and the empty space on the left side of the image. Lines surround the frame within the image.

At least, that's what you'd think... oh, improved, this should work, but not really, so this doesn't improve the image you're going to generate, so we'll continue working and fix it by adding the principle of the Noverra style.

 here comes the prompt according to the Noverra (style > styling) principle: 

Style Fusion

 -Art Nouveau — geometry

- cyberpunk


Color Palette

- Deep Black #000000 — mystery, border

- Muted Gold #D4AF37 — exaltation, symbolism

- Teal #008080 — synthesis, digital clarity

- Midnight Blue #191970 — night, depth, rhythm

- Red #7B1E1E - Ceremony,


Stylized illustration of a woman in profile, A Geometric Art Nouveau fusion. inspired by Art Nouveau and cyberpunk. but with strict geometric framing. No floral motifs, with angular symmetry, no curved lines. Use clean, straight gold lines to frame all four borders — top, bottom, left, and right. The woman is positioned on the right side towards the inside, in a silk red dress, eyes open, facing inward. Background is blue minimal, no organic shapes. no scrollwork. the background is empty.


common mistake with these types of prompts: 

Common mistakes with these types of prompts

  • Using “Art Nouveau” without restriction → leads to curls, flowers, asymmetry
  • “Geometric” without “no curves” → AI thinks: circles are geometric too
  • Too vague instructions such as “clean” or “minimal” → AI fills in what that means itself
  • Color coding without context → AI ignores them unless you tell them where to go
There are still active LORA'S. You better turn them off. Think of other styles and think of curls and not relevant loras. Think of art deco lora while you ask for art nouveau.

a generally good working prompt for this implementation specifically is:

Prompt text for your image (clean geometry, no curls)

General prompt (neutral, suitable for multiple tools):

Stylized illustration of a woman in profile, positioned on the right side of the image, facing inward (toward the left). She has long flowing hair and wears a decorative headband and an ornate dress with embroidery. The background is minimal and dark. Frame the image with thick, straight gold lines on all four borders — top, bottom, left, and right. No circles, no curves, no floral


again the prompt see the differences: this works better 

Frame must be fully visible and aligned within the image boundaries. No cropping or shifting of borders.

Style Fusion
 -Art Nouveau — geometry
- Rococo - gental style of power

Color Palette
- Deep Black #000000 — mystery, border
- Muted Gold #D4AF37 — exaltation, symbolism
- Teal #008080 — synthesis, digital clarity
- Midnight Blue #191970 — night, depth, rhythm
- Red #7B1E1E - Ceremony, glamour 

Stylized illustration of a woman in profile, using a symmetrical and geometrical layout inspired by Art Nouveau and Rococo. She is positioned within the composition on the right side of the image, facing inward (toward the left). No floral motifs. No curved lines. Use clean, straight gold lines to frame all four borders — top, bottom, left, and right. Include both thick and thin line weights. She has long flowing hair and wears a decorative headband and a silk dress with embroidery. The background is minimal and dark. Do not generate any floral elements whatsoever.no curves, no floral, no organic shapes.



quick setting: if you have the setting wide to get the edge:
Setting the Infill Method you're not using yet to — tile — combined with a Tile Size of 32 will definitely affect how sharp and centered your image is. But it can also cause subtle shifts, especially at the edges.

🧠 What happens to “tile” and Tile Size 32?

  • Tile repeats visual patterns in blocks
  • Tile Size 32 means that each block is 32 pixels wide/high
  • When the AI ​​builds the image in this grid structure, the border lines may not align perfectly with the canvas size
  • This can lead to the effect you see: the right and bottom edges appear to fall just outside the frame

✅ What can you do to correct that?

1. Reduce Tile Size to 16 or even 8

  • Smaller blocks = finer grid = better edge alignment

2. Add to your prompt:

“Frame must be fully visible and aligned within the image boundaries. No cropping or shifting of borders.”

3. Combine with tight composition settings:

  • CFG Scale: 9–10
  • Edge Size: 8–12
  • Min Denoise: 0.15
  • Max Denoise: 0.30

: 🧱 Infill Method: Tile — How to prevent edge shifting?

  • Tile Size: 16
  • Prompt line: “Frame must be fully visible…”
  • CFG Scale: 9
  • Edge Size: 8
  • Denoise: 0.15–0.30

Stylized illustration of a woman in profile, using a symmetrical and geometrical layout inspired by Art Nouveau and Rococo. She is positioned within the composition on the right side of the image, facing inward (toward the left). No floral motifs. No curved lines. Use clean, straight gold lines to frame all four borders — top, bottom, left, and right. Include both thick and thin line weights. She has long flowing hair and wears a decorative headband and a silk dress with embroidery. The background is minimal and dark. Do not generate any floral elements whatsoever.no curves, no floral, no organic shapes.Frame must be fully visible and aligned within the image boundaries. No cropping or shifting of borders.”




success 

Links: to the various blog parts.

- The Breachline – Start & Overview


This page is part of the Noverra series, which combines stylistic shifts, rhythm, and digital image structures. Each image carries a unique codex and is aligned with a larger visual system in which form, color, and composition are given meaning. The content is rhythmically placed within the Tussen Klok en Klepel blog universe and aligns with the classification chart and cyclical structure of the series. For more context and in-depth information, visit the Noverra overview page.

Comments

Post a Comment

Bedankt voor uw feedback

Populairste blogs

A new way of thinking about digital art The Noverra-Digital Art style

Image
  What is the Noverra -Digital Art style? A new way of thinking about digital art Noverra is not just a style — it’s a way of working. It’s a hybrid approach to digital art that fuses historical aesthetics with modern technology, symbolic prompts, and intentional color choices to create images that feel layered, ritualistic, and alive. It’s not about perfection or polish. It’s about tension, rhythm, and meaning. What makes Noverra different? Noverra is built on three key elements: 🔸 Style fusion You blend visual styles — like Renaissance, Brutalism, Rococo, and Cyberpunk — not as decoration, but as transformation. The goal isn’t to imitate, but to mutate through contrast. 🔸 Ritual prompt writing Prompts aren’t just instructions. They’re poetic invocations, written with symbolic language to guide the image. You don’t just describe — you evoke. 🔸 Color first In Noverra, color comes before the prompt. You choose a palette that sets the emotional tone — gold for ritua...
Read more »

Reality-tv breekt mensen, en wij kijken toe

Image
Reality-tv en dating shows is niet meer wat het ooit was. Waar het begon als een inkijkje in menselijke relaties, is het nu verworden tot een podium voor publieke afrekening. Deelnemers worden niet alleen beoordeeld door een reflectiepanel — wat op zich nog verdedigbaar is — maar vooral door een publiek dat zich gedraagt alsof het moreel eigenaar is van hun levensverhaal. En dat is een gevaarlijke ontwikkeling. Wat mij het meest stoort, is de manier waarop deelnemers worden aangemoedigd om persoonlijke socialmedia-accounts aan te maken. Dat lijkt een vrije keuze, maar is in werkelijkheid een vooropgezet doel van de makers. Het vergroot de zichtbaarheid, ja — maar ook de kwetsbaarheid. In plaats van bescherming krijgen deelnemers een stortvloed aan meningen, memes en oordelen over zich heen. En die zijn zelden mild. De samenleving kijkt toe, lacht, oordeelt — maar vraagt zich zelden af wat dit met mensen doet. Alsof deelnemers geen echte mensen zijn, maar personages in een fictief scrip...
Read more »

Voor wie denkt, lacht, en af en toe gewoon even niks snapt

Image
  Voor wie denkt, lacht, en af en toe gewoon even niks snapt Een verhaal voor de denkers in badjas, de lachers met vragen, en zoekers zonder routebeschrijving. Er zijn van die dagen dat je hoofd voelt als een rommelzolder. Vol gedachten die je ooit ergens hebt neergelegd, maar waarvan je niet meer weet waarom. Je opent je ogen, je hersenen starten op als een oude laptop die lijkt te draaien op windows 3.0 , en je vraagt je af: Wie ben ik eigenlijk? Niet als naam. Niet als functie. Gewoon als jij, dat rare, denkende wezen dat in een ochtendjasrond strompelt en bijna pijnlijk nadenkt, terwijl de kat op je brood wil liggen. Welkom. Je bent niet alleen. Dit verhaal is voor jou. Voor wie denkt, lacht, en af en toe gewoon even niks snapt. Want soms is dat precies wat je nodig hebt: een beetje verwarring, een beetje herkenning, en een stem die zegt: Je bent oké. Zelfs als je het even niet weet. Wie ben ik als niemand kijkt? Je zit daar. Eén oog half dicht, koffie net sterk genoeg...
Read more »