Images
text-to-image
An image made from nothing but a written description.
<!-- Generated by apps/docs/scripts/gen-tools.mjs. Do not edit here: the words live in apps/web/src/lib/tools/text-to-image.tool.json (agent, hint) and packages/shared/src/tools/setting-hints.ts. --> On the rail as **nano-banana-pro**, **recraft-4**, **recraft-4-flash**, **flux-2-pro**, **flux-2-max**, **qwen-2**, **qwen-3**, **gpt-image-2**, **gpt-image-2.5**, **gpt-image-1.5**. ## When to use it Starting completely from scratch: a scene, a character, a product shot, a texture, or a poster that needs real, readable text on it. ## When to reach for something else It invents everything from the words with no picture to match, so it cannot hold an exact likeness or composition steady between runs without a reference to guide it. ## How to use it 1. Write what you want in **Prompt**. 2. Set **Images** on the card. 3. Pick a model in the pill at the top of the card, or keep **Nano Banana Pro**. 4. Press **Run**. ## On the card | Control | What it does | Choices | | - | - | - | | Prompt | What you want, in your own words. The card sends it with your inputs. | Your words | | Images | How many images to make from this one prompt. Above 1 they are all rendered in a single call. A batch of 8 costs the same as a single image. | 1, 2, 3, 4, 6 or 8 | | Run | Starts the run. The number on it is what this run costs in credits, worked out before you press. | | | ×1 | Runs the card several times in a row with the same settings. | 1 to 99 | ## Models The pill at the top of the card picks the model. **Nano Banana Pro** is the one a new card starts on. | Model | Good for | | - | - | | [Nano Banana Pro](#nano-banana-pro) | The default generator on the text-to-image pill: Google's Nano Banana Pro with flagship prompt adherence and crisp in-image typography, native 1K, 2K or 4K, and a batch count up to eight rendered as one contact sheet for the same credits. | | [Flux 2 Max](#flux-2-max) | Black Forest Labs' highest-fidelity FLUX.2 tier for generation and reference-guided editing: the strongest prompt adherence and most consistent multi-reference editing in the FLUX.2 line, up to six inputs tagged @ in the prompt. | | [Flux 2 Pro](#flux-2-pro) | The FLUX.2 workhorse across three tools: photoreal text-to-image with wide format and aspect control, reference-guided edits with up to six @-tagged inputs, and the engine Recolor uses to repaint a picture to an exact palette. | | [Seedream 5 Pro](#seedream-5-pro) | High-fidelity text-to-image with size presets up to 2K, auto or fixed ratios, from the same Seedream family that edits and composes elsewhere on the canvas. | | [Flux Macro Texture](#flux-macro-texture) | A Flux LoRA fine-tuned on macro photography of surfaces: put the trigger word MCROTX in the prompt and it renders detailed close-up texture images, the texture-plate pick on the text-to-image pill. | | [Recraft v4.1](#recraft-v41) | Art-directed compositions with integrated text, where prompt accuracy and design taste matter: posters, packaging comps, brand visuals; fast and cost-efficient at standard resolution, with an unusually wide ratio list. | | [Recraft v4.1 Flash](#recraft-v41-flash) | Quick, cheap drafts with Recraft design taste: rough poster, packaging or brand directions to compare side by side, a batch of layout ideas, or a picture pushed towards a set palette. | | [Qwen Image 2 Pro](#qwen-image-2-pro) | Text-to-image with legible in-image text, and the same model as an img2img pass with a strength dial from nudge to repaint; it also carries the 2:1 and 1:2 panorama ratios. | | [Qwen Image 3](#qwen-image-3) | Posters, labels and layouts that carry dense, complex typography, which it renders far more accurately than earlier Qwen releases; Chinese and English prompts up to 800 characters, output up to 2048 square. | | [GPT Image 2](#gpt-image-2) | OpenAI GPT Image 2 for prompt-only generation with three quality tiers, wide 21:9 and 3:1 header canvases, output up to 4K, and a batch count of 4, 6 or 8 that renders a contact sheet in one call and slices it into images. | | [GPT 2.5 Flare](#gpt-25-flare) | The everyday GPT Image 2.5 pick: natural light, rich texture and complex layouts from a prompt, five quality steps from low to max, sizes to a 4K pixel budget and wide 21:9 and 3:1 frames; same price as Sunburst and faster. | | [GPT 2.5 Sunburst](#gpt-25-sunburst) | The finer GPT Image 2.5 variant: the same prompt, quality steps and sizes as Flare at the same price, taken when intricate detail matters more than speed. | | [GPT Image 1.5](#gpt-image-15) | OpenAI's GPT Image 1.5, the side-by-side alternative to GPT Image 2 on both the text-to-image and image-edit pills: one model for prompt-only generation and multi-image editing, quality tiered low to high. | | [Ideogram v4](#ideogram-v4) | The typography specialist on the text-to-image pill: posters, logos and signage with legible in-image text, quote the words you want rendered. Speed and quality tiered as Turbo, Balanced and Quality. | | [FLUX 3](#flux-3) | A picture whose composition is decided in advance: a layout of boxes, each element placed where it is drawn, with the words in it spelled exactly, at 1K, 2K or 4K. | ### Nano Banana Pro **Not the one for:** No image input of any kind: editing or referencing a picture is a job for its edit siblings. 4K doubles the price, and a batch above one shares a single sheet, so each image comes back smaller. Behind the settings button: | Setting | What it does | Choices | | - | - | - | | Aspect Ratio | The shape of the result, like 1:1 square, 16:9 wide, or 9:16 tall. "Match input" copies the source shape. | Horizontal 21:9, Horizontal 16:9, Horizontal 3:2, Horizontal 4:3, Horizontal 5:4, Square 1:1, Vertical 4:5, Vertical 3:4, Vertical 2:3, Vertical 9:16 | | Resolution | Batches render at 2K. 1K and 2K cost the same on this model. | 1K, 2K, 4K | | Seed | The starting dice roll. The same seed and settings give the same result every time; change it for a fresh variation, or leave it at -1 for a new random one. | -1 to 999999 | ### Flux 2 Max **Not the one for:** Costs about twice Pro per megapixel for the same request shape, so when the Pro result already reads right, Max is the pricier route. No mask input, and output above 2 MP is not recommended. Behind the settings button: | Setting | What it does | Choices | | - | - | - | | Aspect Ratio | The shape of the result, like 1:1 square, 16:9 wide, or 9:16 tall. "Match input" copies the source shape. | Square 1:1, Horizontal 16:9, Vertical 9:16, Horizontal 4:3, Vertical 3:4, Horizontal 3:2, Vertical 2:3, Vertical 4:5, Horizontal 5:4 | | Resolution | How big and detailed the result is. Higher looks sharper but costs more and takes longer. | 0.5 MP, 1 MP, 2 MP, 4 MP | | Seed | The starting dice roll. The same seed and settings give the same result every time; change it for a fresh variation, or leave it at -1 for a new random one. | -1 to 999999 | | Output Format | The file type: PNG keeps the best quality and supports transparency, JPG is a smaller file, WebP is small and modern. | PNG, WebP, JPG | ### Flux 2 Pro **Not the one for:** No mask, so a change confined to one painted region belongs to inpainting, and it re-composes freely from references rather than holding a subject at its pixel position. Output above 2 MP is not recommended. Behind the settings button: | Setting | What it does | Choices | | - | - | - | | Aspect Ratio | The shape of the result, like 1:1 square, 16:9 wide, or 9:16 tall. "Match input" copies the source shape. | Square 1:1, Horizontal 16:9, Vertical 9:16, Horizontal 4:3, Vertical 3:4, Horizontal 3:2, Vertical 2:3, Vertical 4:5, Horizontal 5:4 | | Resolution | How big and detailed the result is. Higher looks sharper but costs more and takes longer. | 0.5 MP, 1 MP, 2 MP, 4 MP | | Seed | The starting dice roll. The same seed and settings give the same result every time; change it for a fresh variation, or leave it at -1 for a new random one. | -1 to 999999 | | Output Format | The file type: PNG keeps the best quality and supports transparency, JPG is a smaller file, WebP is small and modern. | PNG, WebP, JPG | ### Seedream 5 Pro **Not the one for:** Output stops at the 2K preset and there is no seed, quality ladder or negative prompt, so a 4K deliverable or a repeatable seed-locked series needs a different model. Behind the settings button: | Setting | What it does | Choices | | - | - | - | | Image Size | The size and shape of the image. Bigger looks sharper but costs more and is slower. | Auto 2K, Auto 1K, Square small, Square large, Horizontal 16:9, Vertical 9:16, Horizontal 4:3, Vertical 3:4 | ### Flux Macro Texture **Not the one for:** Single-purpose: not for scenes, characters, products or in-image text, and it takes no image input. Without the MCROTX trigger word the LoRA style does not fire. Behind the settings button: | Setting | What it does | Choices | | - | - | - | | Aspect Ratio | The shape of the result, like 1:1 square, 16:9 wide, or 9:16 tall. "Match input" copies the source shape. | Square 1:1, Horizontal 21:9, Horizontal 16:9, Horizontal 3:2, Horizontal 4:3, Horizontal 5:4, Vertical 4:5, Vertical 3:4, Vertical 2:3, Vertical 9:16, Vertical 9:21 | | Steps | How many passes the tool takes to refine the result. More looks a bit smoother but is slower. | 10 to 50 | | Guidance Scale | How tightly the result sticks to your prompt. Higher follows your words more literally; lower gives the tool more freedom. | 1 to 20 | | Seed | The starting dice roll. The same seed and settings give the same result every time; change it for a fresh variation, or leave it at -1 for a new random one. | -1 to 999999 | ### Recraft v4.1 **Not the one for:** Sizes stop around 1536px on the long edge and there is no reference-image input, so an edit of an existing picture or a native 2K or 4K output goes elsewhere. Behind the settings button: | Setting | What it does | Choices | | - | - | - | | Aspect Ratio | The shape of the result, like 1:1 square, 16:9 wide, or 9:16 tall. "Match input" copies the source shape. | Not set, Square 1:1, Horizontal 4:3, Vertical 3:4, Horizontal 3:2, Vertical 2:3, Horizontal 16:9, Vertical 9:16, Vertical 4:5, Horizontal 5:4, Horizontal 2:1, Vertical 1:2, Horizontal 14:10, Vertical 10:14, Vertical 6:10 | | Size | The engine size: Large is the highest quality, Turbo is the fastest and cheapest. | 1024×1024, 1536×768, 768×1536, 1280×832, 832×1280, 1216×896, 896×1216, 1152×896, 896×1152, 832×1344, 1280×896, 896×1280, 1344×768, 768×1344 | ### Recraft v4.1 Flash **Not the one for:** It is the fast tier, so detail and text rendering sit below the full Recraft model; there is no reference-image input and every picture is about one megapixel (1536 pixels at most on the long edge). Behind the settings button: | Setting | What it does | Choices | | - | - | - | | Aspect Ratio | The shape of the picture. Every shape comes out at about one megapixel and costs the same. | Square 1:1, Horizontal 4:3, Vertical 3:4, Horizontal 3:2, Vertical 2:3, Horizontal 16:9, Vertical 9:16, Vertical 4:5, Horizontal 5:4, Horizontal 2:1, Vertical 1:2, Horizontal 14:10, Vertical 10:14, Vertical 6:10 | ### Qwen Image 2 Pro **Not the one for:** Takes at most one reference image and has no mask, so multi-image blends and strictly local fixes belong elsewhere; the strength slider is the only edit control. Behind the settings button: | Setting | What it does | Choices | | - | - | - | | Aspect Ratio | The shape of the result, like 1:1 square, 16:9 wide, or 9:16 tall. "Match input" copies the source shape. | Square 1:1, Horizontal 16:9, Vertical 9:16, Horizontal 4:3, Vertical 3:4, Horizontal 3:2, Vertical 2:3, Horizontal 2:1, Vertical 1:2 | ### Qwen Image 3 **Not the one for:** A run takes around two minutes and the prompt is capped at 800 characters, so it is the wrong pick for a fast batch of rough ideas or for a very long brief. Behind the settings button: | Setting | What it does | Choices | | - | - | - | | Image Size | The size and shape of the image. Bigger looks sharper but costs more and is slower. | Square large, Square small, Horizontal 16:9, Vertical 9:16, Horizontal 4:3, Vertical 3:4 | | Prompt Expansion | Quietly fleshes out a short prompt with extra detail before generating. | On or off | ### GPT Image 2 **Not the one for:** No image input of any kind; editing a picture is its edit sibling's job. A batch renders at 2K and is charged for it, so four cells cost more than one 1K image, and the high tier is the expensive one. Behind the settings button: | Setting | What it does | Choices | | - | - | - | | Aspect Ratio | The shape of the picture. 21:9 and 3:1 are header shapes. They render on a custom canvas at or below the pixel count of Square large, so they cost the same as every other size here. | Square large, Square small, Vertical 3:4, Vertical 9:16, Horizontal 4:3, Horizontal 16:9, Horizontal 21:9, Horizontal 3:1 | | Quality | The overall finish level. Higher looks better but costs more and is slower. | Low, Medium, High | | Size | Output resolution. 4K fills an 8.3 megapixel budget, so a 16:9 image comes back 3840x2160 and a square one 2880x2880. Asking for several images has always rendered the sheet at 2K, so 1K there still gives you 2K. | Standard, Large, 4K | ### GPT 2.5 Flare **Not the one for:** It takes no reference image, so anything starting from an existing picture belongs with its edit sibling; and 4K here is a pixel budget, not an edge, so a square 4K render is 2880px rather than 4096. Behind the settings button: | Setting | What it does | Choices | | - | - | - | | Aspect Ratio | The shape of the picture. 21:9 and 3:1 are header shapes. They render on a custom canvas at or below the pixel count of Square large, so they cost the same as every other size here. | Square large, Square small, Vertical 3:4, Vertical 9:16, Horizontal 4:3, Horizontal 16:9, Horizontal 21:9, Horizontal 3:1 | | Quality | How hard the model works on the picture. Higher looks better, costs more and takes longer (1, 2, 9, 13 and 32 credits for a square picture at Standard size). | Low, Medium, High, Very High, Max | | Size | Output resolution. 4K fills an 8.3 megapixel budget, so a 16:9 image comes back 3840x2160 and a square one 2880x2880. Asking for several images has always rendered the sheet at 2K, so 1K there still gives you 2K. | Standard, Large, 4K | ### GPT 2.5 Sunburst **Not the one for:** Roughly twice as slow as Flare above high quality at an identical price, so it is the wrong pick when turnaround matters more than intricate detail. Behind the settings button: | Setting | What it does | Choices | | - | - | - | | Aspect Ratio | The shape of the picture. 21:9 and 3:1 are header shapes. They render on a custom canvas at or below the pixel count of Square large, so they cost the same as every other size here. | Square large, Square small, Vertical 3:4, Vertical 9:16, Horizontal 4:3, Horizontal 16:9, Horizontal 21:9, Horizontal 3:1 | | Quality | How hard the model works on the picture. Higher looks better, costs more and takes longer (1, 2, 9, 13 and 32 credits for a square picture at Standard size). | Low, Medium, High, Very High, Max | | Size | Output resolution. 4K fills an 8.3 megapixel budget, so a 16:9 image comes back 3840x2160 and a square one 2880x2880. Asking for several images has always rendered the sheet at 2K, so 1K there still gives you 2K. | Standard, Large, 4K | ### GPT Image 1.5 **Not the one for:** Only three aspect ratios (1:1, 3:2, 2:3), no mask input and no batch count. The high tier is the expensive one; the image-edit pill opens it at medium and text-to-image at high. Behind the settings button: | Setting | What it does | Choices | | - | - | - | | Quality | The overall finish level. Higher looks better but costs more and is slower. | Low, Medium, High | | Aspect Ratio | The shape of the result, like 1:1 square, 16:9 wide, or 9:16 tall. "Match input" copies the source shape. | Square 1:1, Horizontal 3:2, Vertical 2:3 | ### Ideogram v4 **Not the one for:** Text-only, no image input and one image per run. Cost scales with output megapixels and tier, and prompt expansion adds a small per-image cost when it is left on. Behind the settings button: | Setting | What it does | Choices | | - | - | - | | Rendering Speed | Trade speed, finish, and cost: Turbo is fastest and cheapest, Quality looks best but costs the most. | Turbo, Balanced, Quality | | Image Size | The size and shape of the image. Bigger looks sharper but costs more and is slower. | Square large, Square small, Vertical 3:4, Vertical 9:16, Horizontal 4:3, Horizontal 16:9 | | Acceleration | A speed-up setting. Higher is faster but can slightly soften the result. | None, Low, Regular, High | | Output Format | The file type: PNG keeps the best quality and supports transparency, JPG is a smaller file, WebP is small and modern. | PNG, JPEG | | Safety Checker | Filters out unsafe content when on. | On or off | | Seed | The starting dice roll. The same seed and settings give the same result every time; change it for a fresh variation, or leave it at -1 for a new random one. | -1 to 999999 | | Aesthetics | Optional words on the overall look, like a bold modern poster. | Your words | | Lighting | Optional words on the light, like a hard key light from the upper left. | Your words | | Art style | Optional words on the medium, like flat vector or 3D rendered. | Your words | ### FLUX 3 **Not the one for:** No seed, so the same layout comes back different each run, and one picture a run. 4K is the expensive tier and the slow one. Behind the settings button: | Setting | What it does | Choices | | - | - | - | | Aspect Ratio | The shape of the result, like 1:1 square, 16:9 wide, or 9:16 tall. "Match input" copies the source shape. | Square 1:1, Vertical 2:3, Horizontal 3:2, Vertical 3:4, Horizontal 4:3, Vertical 4:5, Horizontal 5:4, Vertical 9:16, Horizontal 16:9, Horizontal 21:9 | | Resolution | How big and detailed the result is. Higher looks sharper but costs more and takes longer. | 1K, 2K, 4K |On the rail as nano-banana-pro, recraft-4, recraft-4-flash, flux-2-pro, flux-2-max, qwen-2, qwen-3, gpt-image-2, gpt-image-2.5, gpt-image-1.5.
When to use it
Section titled “When to use it”Starting completely from scratch: a scene, a character, a product shot, a texture, or a poster that needs real, readable text on it.
When to reach for something else
Section titled “When to reach for something else”It invents everything from the words with no picture to match, so it cannot hold an exact likeness or composition steady between runs without a reference to guide it.
How to use it
Section titled “How to use it”- Write what you want in Prompt.
- Set Images on the card.
- Pick a model in the pill at the top of the card, or keep Nano Banana Pro.
- Press Run.
On the card
Section titled “On the card”| Control | What it does | Choices |
|---|---|---|
| Prompt | What you want, in your own words. The card sends it with your inputs. | Your words |
| Images | How many images to make from this one prompt. Above 1 they are all rendered in a single call. A batch of 8 costs the same as a single image. | 1, 2, 3, 4, 6 or 8 |
| Run | Starts the run. The number on it is what this run costs in credits, worked out before you press. | |
| ×1 | Runs the card several times in a row with the same settings. | 1 to 99 |
Models
Section titled “Models”The pill at the top of the card picks the model. Nano Banana Pro is the one a new card starts on.
| Model | Good for |
|---|---|
| Nano Banana Pro | The default generator on the text-to-image pill: Google’s Nano Banana Pro with flagship prompt adherence and crisp in-image typography, native 1K, 2K or 4K, and a batch count up to eight rendered as one contact sheet for the same credits. |
| Flux 2 Max | Black Forest Labs’ highest-fidelity FLUX.2 tier for generation and reference-guided editing: the strongest prompt adherence and most consistent multi-reference editing in the FLUX.2 line, up to six inputs tagged @ in the prompt. |
| Flux 2 Pro | The FLUX.2 workhorse across three tools: photoreal text-to-image with wide format and aspect control, reference-guided edits with up to six @-tagged inputs, and the engine Recolor uses to repaint a picture to an exact palette. |
| Seedream 5 Pro | High-fidelity text-to-image with size presets up to 2K, auto or fixed ratios, from the same Seedream family that edits and composes elsewhere on the canvas. |
| Flux Macro Texture | A Flux LoRA fine-tuned on macro photography of surfaces: put the trigger word MCROTX in the prompt and it renders detailed close-up texture images, the texture-plate pick on the text-to-image pill. |
| Recraft v4.1 | Art-directed compositions with integrated text, where prompt accuracy and design taste matter: posters, packaging comps, brand visuals; fast and cost-efficient at standard resolution, with an unusually wide ratio list. |
| Recraft v4.1 Flash | Quick, cheap drafts with Recraft design taste: rough poster, packaging or brand directions to compare side by side, a batch of layout ideas, or a picture pushed towards a set palette. |
| Qwen Image 2 Pro | Text-to-image with legible in-image text, and the same model as an img2img pass with a strength dial from nudge to repaint; it also carries the 2:1 and 1:2 panorama ratios. |
| Qwen Image 3 | Posters, labels and layouts that carry dense, complex typography, which it renders far more accurately than earlier Qwen releases; Chinese and English prompts up to 800 characters, output up to 2048 square. |
| GPT Image 2 | OpenAI GPT Image 2 for prompt-only generation with three quality tiers, wide 21:9 and 3:1 header canvases, output up to 4K, and a batch count of 4, 6 or 8 that renders a contact sheet in one call and slices it into images. |
| GPT 2.5 Flare | The everyday GPT Image 2.5 pick: natural light, rich texture and complex layouts from a prompt, five quality steps from low to max, sizes to a 4K pixel budget and wide 21:9 and 3:1 frames; same price as Sunburst and faster. |
| GPT 2.5 Sunburst | The finer GPT Image 2.5 variant: the same prompt, quality steps and sizes as Flare at the same price, taken when intricate detail matters more than speed. |
| GPT Image 1.5 | OpenAI’s GPT Image 1.5, the side-by-side alternative to GPT Image 2 on both the text-to-image and image-edit pills: one model for prompt-only generation and multi-image editing, quality tiered low to high. |
| Ideogram v4 | The typography specialist on the text-to-image pill: posters, logos and signage with legible in-image text, quote the words you want rendered. Speed and quality tiered as Turbo, Balanced and Quality. |
| FLUX 3 | A picture whose composition is decided in advance: a layout of boxes, each element placed where it is drawn, with the words in it spelled exactly, at 1K, 2K or 4K. |
Nano Banana Pro
Section titled “Nano Banana Pro”Not the one for: No image input of any kind: editing or referencing a picture is a job for its edit siblings. 4K doubles the price, and a batch above one shares a single sheet, so each image comes back smaller.
Behind the settings button:
| Setting | What it does | Choices |
|---|---|---|
| Aspect Ratio | The shape of the result, like 1:1 square, 16:9 wide, or 9:16 tall. “Match input” copies the source shape. | Horizontal 21:9, Horizontal 16:9, Horizontal 3:2, Horizontal 4:3, Horizontal 5:4, Square 1:1, Vertical 4:5, Vertical 3:4, Vertical 2:3, Vertical 9:16 |
| Resolution | Batches render at 2K. 1K and 2K cost the same on this model. | 1K, 2K, 4K |
| Seed | The starting dice roll. The same seed and settings give the same result every time; change it for a fresh variation, or leave it at -1 for a new random one. | -1 to 999999 |
Flux 2 Max
Section titled “Flux 2 Max”Not the one for: Costs about twice Pro per megapixel for the same request shape, so when the Pro result already reads right, Max is the pricier route. No mask input, and output above 2 MP is not recommended.
Behind the settings button:
| Setting | What it does | Choices |
|---|---|---|
| Aspect Ratio | The shape of the result, like 1:1 square, 16:9 wide, or 9:16 tall. “Match input” copies the source shape. | Square 1:1, Horizontal 16:9, Vertical 9:16, Horizontal 4:3, Vertical 3:4, Horizontal 3:2, Vertical 2:3, Vertical 4:5, Horizontal 5:4 |
| Resolution | How big and detailed the result is. Higher looks sharper but costs more and takes longer. | 0.5 MP, 1 MP, 2 MP, 4 MP |
| Seed | The starting dice roll. The same seed and settings give the same result every time; change it for a fresh variation, or leave it at -1 for a new random one. | -1 to 999999 |
| Output Format | The file type: PNG keeps the best quality and supports transparency, JPG is a smaller file, WebP is small and modern. | PNG, WebP, JPG |
Flux 2 Pro
Section titled “Flux 2 Pro”Not the one for: No mask, so a change confined to one painted region belongs to inpainting, and it re-composes freely from references rather than holding a subject at its pixel position. Output above 2 MP is not recommended.
Behind the settings button:
| Setting | What it does | Choices |
|---|---|---|
| Aspect Ratio | The shape of the result, like 1:1 square, 16:9 wide, or 9:16 tall. “Match input” copies the source shape. | Square 1:1, Horizontal 16:9, Vertical 9:16, Horizontal 4:3, Vertical 3:4, Horizontal 3:2, Vertical 2:3, Vertical 4:5, Horizontal 5:4 |
| Resolution | How big and detailed the result is. Higher looks sharper but costs more and takes longer. | 0.5 MP, 1 MP, 2 MP, 4 MP |
| Seed | The starting dice roll. The same seed and settings give the same result every time; change it for a fresh variation, or leave it at -1 for a new random one. | -1 to 999999 |
| Output Format | The file type: PNG keeps the best quality and supports transparency, JPG is a smaller file, WebP is small and modern. | PNG, WebP, JPG |
Seedream 5 Pro
Section titled “Seedream 5 Pro”Not the one for: Output stops at the 2K preset and there is no seed, quality ladder or negative prompt, so a 4K deliverable or a repeatable seed-locked series needs a different model.
Behind the settings button:
| Setting | What it does | Choices |
|---|---|---|
| Image Size | The size and shape of the image. Bigger looks sharper but costs more and is slower. | Auto 2K, Auto 1K, Square small, Square large, Horizontal 16:9, Vertical 9:16, Horizontal 4:3, Vertical 3:4 |
Flux Macro Texture
Section titled “Flux Macro Texture”Not the one for: Single-purpose: not for scenes, characters, products or in-image text, and it takes no image input. Without the MCROTX trigger word the LoRA style does not fire.
Behind the settings button:
| Setting | What it does | Choices |
|---|---|---|
| Aspect Ratio | The shape of the result, like 1:1 square, 16:9 wide, or 9:16 tall. “Match input” copies the source shape. | Square 1:1, Horizontal 21:9, Horizontal 16:9, Horizontal 3:2, Horizontal 4:3, Horizontal 5:4, Vertical 4:5, Vertical 3:4, Vertical 2:3, Vertical 9:16, Vertical 9:21 |
| Steps | How many passes the tool takes to refine the result. More looks a bit smoother but is slower. | 10 to 50 |
| Guidance Scale | How tightly the result sticks to your prompt. Higher follows your words more literally; lower gives the tool more freedom. | 1 to 20 |
| Seed | The starting dice roll. The same seed and settings give the same result every time; change it for a fresh variation, or leave it at -1 for a new random one. | -1 to 999999 |
Recraft v4.1
Section titled “Recraft v4.1”Not the one for: Sizes stop around 1536px on the long edge and there is no reference-image input, so an edit of an existing picture or a native 2K or 4K output goes elsewhere.
Behind the settings button:
| Setting | What it does | Choices |
|---|---|---|
| Aspect Ratio | The shape of the result, like 1:1 square, 16:9 wide, or 9:16 tall. “Match input” copies the source shape. | Not set, Square 1:1, Horizontal 4:3, Vertical 3:4, Horizontal 3:2, Vertical 2:3, Horizontal 16:9, Vertical 9:16, Vertical 4:5, Horizontal 5:4, Horizontal 2:1, Vertical 1:2, Horizontal 14:10, Vertical 10:14, Vertical 6:10 |
| Size | The engine size: Large is the highest quality, Turbo is the fastest and cheapest. | 1024×1024, 1536×768, 768×1536, 1280×832, 832×1280, 1216×896, 896×1216, 1152×896, 896×1152, 832×1344, 1280×896, 896×1280, 1344×768, 768×1344 |
Recraft v4.1 Flash
Section titled “Recraft v4.1 Flash”Not the one for: It is the fast tier, so detail and text rendering sit below the full Recraft model; there is no reference-image input and every picture is about one megapixel (1536 pixels at most on the long edge).
Behind the settings button:
| Setting | What it does | Choices |
|---|---|---|
| Aspect Ratio | The shape of the picture. Every shape comes out at about one megapixel and costs the same. | Square 1:1, Horizontal 4:3, Vertical 3:4, Horizontal 3:2, Vertical 2:3, Horizontal 16:9, Vertical 9:16, Vertical 4:5, Horizontal 5:4, Horizontal 2:1, Vertical 1:2, Horizontal 14:10, Vertical 10:14, Vertical 6:10 |
Qwen Image 2 Pro
Section titled “Qwen Image 2 Pro”Not the one for: Takes at most one reference image and has no mask, so multi-image blends and strictly local fixes belong elsewhere; the strength slider is the only edit control.
Behind the settings button:
| Setting | What it does | Choices |
|---|---|---|
| Aspect Ratio | The shape of the result, like 1:1 square, 16:9 wide, or 9:16 tall. “Match input” copies the source shape. | Square 1:1, Horizontal 16:9, Vertical 9:16, Horizontal 4:3, Vertical 3:4, Horizontal 3:2, Vertical 2:3, Horizontal 2:1, Vertical 1:2 |
Qwen Image 3
Section titled “Qwen Image 3”Not the one for: A run takes around two minutes and the prompt is capped at 800 characters, so it is the wrong pick for a fast batch of rough ideas or for a very long brief.
Behind the settings button:
| Setting | What it does | Choices |
|---|---|---|
| Image Size | The size and shape of the image. Bigger looks sharper but costs more and is slower. | Square large, Square small, Horizontal 16:9, Vertical 9:16, Horizontal 4:3, Vertical 3:4 |
| Prompt Expansion | Quietly fleshes out a short prompt with extra detail before generating. | On or off |
GPT Image 2
Section titled “GPT Image 2”Not the one for: No image input of any kind; editing a picture is its edit sibling’s job. A batch renders at 2K and is charged for it, so four cells cost more than one 1K image, and the high tier is the expensive one.
Behind the settings button:
| Setting | What it does | Choices |
|---|---|---|
| Aspect Ratio | The shape of the picture. 21:9 and 3:1 are header shapes. They render on a custom canvas at or below the pixel count of Square large, so they cost the same as every other size here. | Square large, Square small, Vertical 3:4, Vertical 9:16, Horizontal 4:3, Horizontal 16:9, Horizontal 21:9, Horizontal 3:1 |
| Quality | The overall finish level. Higher looks better but costs more and is slower. | Low, Medium, High |
| Size | Output resolution. 4K fills an 8.3 megapixel budget, so a 16:9 image comes back 3840x2160 and a square one 2880x2880. Asking for several images has always rendered the sheet at 2K, so 1K there still gives you 2K. | Standard, Large, 4K |
GPT 2.5 Flare
Section titled “GPT 2.5 Flare”Not the one for: It takes no reference image, so anything starting from an existing picture belongs with its edit sibling; and 4K here is a pixel budget, not an edge, so a square 4K render is 2880px rather than 4096.
Behind the settings button:
| Setting | What it does | Choices |
|---|---|---|
| Aspect Ratio | The shape of the picture. 21:9 and 3:1 are header shapes. They render on a custom canvas at or below the pixel count of Square large, so they cost the same as every other size here. | Square large, Square small, Vertical 3:4, Vertical 9:16, Horizontal 4:3, Horizontal 16:9, Horizontal 21:9, Horizontal 3:1 |
| Quality | How hard the model works on the picture. Higher looks better, costs more and takes longer (1, 2, 9, 13 and 32 credits for a square picture at Standard size). | Low, Medium, High, Very High, Max |
| Size | Output resolution. 4K fills an 8.3 megapixel budget, so a 16:9 image comes back 3840x2160 and a square one 2880x2880. Asking for several images has always rendered the sheet at 2K, so 1K there still gives you 2K. | Standard, Large, 4K |
GPT 2.5 Sunburst
Section titled “GPT 2.5 Sunburst”Not the one for: Roughly twice as slow as Flare above high quality at an identical price, so it is the wrong pick when turnaround matters more than intricate detail.
Behind the settings button:
| Setting | What it does | Choices |
|---|---|---|
| Aspect Ratio | The shape of the picture. 21:9 and 3:1 are header shapes. They render on a custom canvas at or below the pixel count of Square large, so they cost the same as every other size here. | Square large, Square small, Vertical 3:4, Vertical 9:16, Horizontal 4:3, Horizontal 16:9, Horizontal 21:9, Horizontal 3:1 |
| Quality | How hard the model works on the picture. Higher looks better, costs more and takes longer (1, 2, 9, 13 and 32 credits for a square picture at Standard size). | Low, Medium, High, Very High, Max |
| Size | Output resolution. 4K fills an 8.3 megapixel budget, so a 16:9 image comes back 3840x2160 and a square one 2880x2880. Asking for several images has always rendered the sheet at 2K, so 1K there still gives you 2K. | Standard, Large, 4K |
GPT Image 1.5
Section titled “GPT Image 1.5”Not the one for: Only three aspect ratios (1:1, 3:2, 2:3), no mask input and no batch count. The high tier is the expensive one; the image-edit pill opens it at medium and text-to-image at high.
Behind the settings button:
| Setting | What it does | Choices |
|---|---|---|
| Quality | The overall finish level. Higher looks better but costs more and is slower. | Low, Medium, High |
| Aspect Ratio | The shape of the result, like 1:1 square, 16:9 wide, or 9:16 tall. “Match input” copies the source shape. | Square 1:1, Horizontal 3:2, Vertical 2:3 |
Ideogram v4
Section titled “Ideogram v4”Not the one for: Text-only, no image input and one image per run. Cost scales with output megapixels and tier, and prompt expansion adds a small per-image cost when it is left on.
Behind the settings button:
| Setting | What it does | Choices |
|---|---|---|
| Rendering Speed | Trade speed, finish, and cost: Turbo is fastest and cheapest, Quality looks best but costs the most. | Turbo, Balanced, Quality |
| Image Size | The size and shape of the image. Bigger looks sharper but costs more and is slower. | Square large, Square small, Vertical 3:4, Vertical 9:16, Horizontal 4:3, Horizontal 16:9 |
| Acceleration | A speed-up setting. Higher is faster but can slightly soften the result. | None, Low, Regular, High |
| Output Format | The file type: PNG keeps the best quality and supports transparency, JPG is a smaller file, WebP is small and modern. | PNG, JPEG |
| Safety Checker | Filters out unsafe content when on. | On or off |
| Seed | The starting dice roll. The same seed and settings give the same result every time; change it for a fresh variation, or leave it at -1 for a new random one. | -1 to 999999 |
| Aesthetics | Optional words on the overall look, like a bold modern poster. | Your words |
| Lighting | Optional words on the light, like a hard key light from the upper left. | Your words |
| Art style | Optional words on the medium, like flat vector or 3D rendered. | Your words |
FLUX 3
Section titled “FLUX 3”Not the one for: No seed, so the same layout comes back different each run, and one picture a run. 4K is the expensive tier and the slow one.
Behind the settings button:
| Setting | What it does | Choices |
|---|---|---|
| Aspect Ratio | The shape of the result, like 1:1 square, 16:9 wide, or 9:16 tall. “Match input” copies the source shape. | Square 1:1, Vertical 2:3, Horizontal 3:2, Vertical 3:4, Horizontal 4:3, Vertical 4:5, Horizontal 5:4, Vertical 9:16, Horizontal 16:9, Horizontal 21:9 |
| Resolution | How big and detailed the result is. Higher looks sharper but costs more and takes longer. | 1K, 2K, 4K |