Private Diffusion # AI Image Prompt Generator Build a prompt one word at a time and see what each word changes, on the same seed. ![Fox: a fox in a misty forest](/_astro/base.BRUN2IPd_ZqMFkc.jpg) ![Portrait: photorealistic portrait of a blonde woman with blue eyes](/_astro/base.De_QfOdv_1RBSoO.jpg) ![Still life: a ceramic teapot on a wooden table](/_astro/base.CgdUZoJQ_Z17i6Rj.jpg) Example subjectFoxPortraitStill life Hold to see without it **Subject only**The example subject with nothing added. Same seed, one word changed · Made with [Kroma Turbo](/models/kroma-turbo) on a 16 GB iPad or Mac Renders show the fox. Your prompt \+ action\+ setting\+ camera\+ light\+ style a fox in a misty forest Edit as textBack to builderTiles and subject are paused while you edit. Writing for ![](/model-logos/circlestone-labs.png)Anima![](/model-logos/black-forest-labs.svg)Flux.2 Klein 4BMFMage Flow Turbo![](/model-logos/krea.svg)Krea 2 Turbo![](/model-logos/lodestones.webp)Kroma Turbo+5 more ![](/model-logos/tongyi-lab.png)Z-Image-Turbo![](/model-logos/rundiffusion.png)Juggernaut Z Fast![](/model-logos/baidu.webp)ERNIE Image Turbo![](/model-logos/lodestones.webp)Chroma1-Flash![](/model-logos/boogu.png)Boogu Image Turbo PartActionSettingCameraLightStyleShuffle Choosing a word needs JavaScript. The pictures still show what each word does. ![](/_astro/action.facing-camera.992lrMFh_2aWSW3.jpg)Facing the camera![](/_astro/action.in-profile.B7KSMuPl_Sj2Ds.jpg)In profile![](/_astro/action.from-behind.DkKSx_Y8_Z1utnHK.jpg)Seen from behind![](/_astro/action.wind-blown.HicxGlQo_Z1Oyw96.jpg)Caught in the wind![](/_astro/action.in-motion.BpHT6KHm_XYfCu.jpg)In motion![](/_astro/action.at-rest.DR1LcNMU_9sd6a.jpg)At rest![](/_astro/action.partly-hidden.o9Ik1bWm_Z1BzFaU.jpg)Partly hidden![](/_astro/action.steam-rising.Bdl0xMDY_1jrLDo.jpg)Steam risingShuffle ![](/_astro/setting.studio-backdrop.XY4tGa9r_Z21IWqc.jpg)Plain studio backdrop![](/_astro/setting.city-street.CpKBPAuC_ZGgMrJ.jpg)City street![](/_astro/setting.beach.CYYVhhpo_Z1couhH.jpg)Beach![](/_astro/setting.fresh-snow.CZ98378Z_Z2wazYN.jpg)Fresh snow![](/_astro/setting.cozy-kitchen.jM9v4sBv_1IbWmd.jpg)Cozy kitchen![](/_astro/setting.old-library.CVDi-LLo_2u3lgv.jpg)Old library![](/_astro/setting.mountain-meadow.XrDLYh6D_Z1cp9eD.jpg)Mountain meadow![](/_astro/setting.light-rain.85pwMj43_1hCyhS.jpg)Light rainShuffle ![](/_astro/camera.close-up.Spu7crGM_ZsBVi5.jpg)Close-up![](/_astro/camera.wide-shot.DKP3wxhj_20RtC0.jpg)Wide shot![](/_astro/camera.low-angle.DtKzbtAt_1TCUih.jpg)Low angle![](/_astro/camera.overhead.BMIMGrYe_zVgd.jpg)Overhead![](/_astro/camera.shallow-focus.DrlCvgMj_Z5ti23.jpg)Shallow focus![](/_astro/camera.telephoto.DvSSHqV__27gW0E.jpg)Telephoto lens![](/_astro/camera.wide-angle.B6Sfcmd__6xezI.jpg)Wide-angle lens![](/_astro/camera.macro.DXGZxT_8_2vrebi.jpg)MacroShuffle ![](/_astro/light.window.C_smPZmt_Z2b9AtU.jpg)Window light![](/_astro/light.golden-hour.Cea7xM5E_ZPpcVm.jpg)Golden hour![](/_astro/light.overcast.Cwqcn6hB_2dwXh4.jpg)Overcast![](/_astro/light.backlit.Cks0acrr_gkfwI.jpg)Backlit![](/_astro/light.candlelight.CFvOEr23_Zem4xC.jpg)Candlelight![](/_astro/light.neon.BwumFYCr_Lsu24.jpg)Neon![](/_astro/light.studio-softbox.i9fsC8uq_Z1jtHyS.jpg)Studio softbox![](/_astro/light.moonlight.BYQbreww_ZFNyMU.jpg)MoonlightShuffle ![](/_astro/style.film-photo.DoQJboCJ_Z1YH3VG.jpg)35mm film![](/_astro/style.watercolor.DPW-cvt6_ZBhvN8.jpg)Watercolor![](/_astro/style.oil-painting.WZin1-NH_Z1ncf52.jpg)Oil painting![](/_astro/style.anime.BlzoQbue_2prSzC.jpg)Anime![](/_astro/style.black-and-white.CVwHHMjP_UR3EH.jpg)Black and white![](/_astro/style.3d-render.BGnfthgD_ZoN8eo.jpg)3D render![](/_astro/style.pencil-sketch.Dq_Kvyjf_1gcPSH.jpg)Pencil sketch![](/_astro/style.cinematic.C_SHilm-_1d100j.jpg)Cinematic stillShuffle Copy promptShare For Studio AspectVertical 9:16Portrait 2:3Portrait 4:5Square 1:1Landscape 3:2Wide 16:9Cine 21:9 Negative promptCopy negative prompt Size S · M · L everywhere; XL on 24 GB devices Rewrite Prompt This model has no rewrite step Composing… Retry ![](/_astro/action.wind-blown.HicxGlQo_Z1Oyw96.jpg)![](/_astro/setting.mountain-meadow.XrDLYh6D_Z1cp9eD.jpg)![](/_astro/camera.telephoto.DvSSHqV__27gW0E.jpg)![](/_astro/light.golden-hour.Cea7xM5E_ZPpcVm.jpg)![](/_astro/style.film-photo.DoQJboCJ_Z1YH3VG.jpg)Golden Hour Film![](/_astro/action.in-motion.BpHT6KHm_XYfCu.jpg)![](/_astro/setting.city-street.CpKBPAuC_ZGgMrJ.jpg)![](/_astro/camera.wide-angle.B6Sfcmd__6xezI.jpg)![](/_astro/light.neon.BwumFYCr_Lsu24.jpg)![](/_astro/style.cinematic.C_SHilm-_1d100j.jpg)Neon Street![](/_astro/action.at-rest.DR1LcNMU_9sd6a.jpg)![](/_astro/setting.old-library.CVDi-LLo_2u3lgv.jpg)![](/_astro/camera.shallow-focus.DrlCvgMj_Z5ti23.jpg)![](/_astro/light.window.C_smPZmt_Z2b9AtU.jpg)![](/_astro/style.oil-painting.WZin1-NH_Z1ncf52.jpg)Quiet Study![](/_astro/action.partly-hidden.o9Ik1bWm_Z1BzFaU.jpg)![](/_astro/setting.fresh-snow.CZ98378Z_Z2wazYN.jpg)![](/_astro/camera.close-up.Spu7crGM_ZsBVi5.jpg)![](/_astro/light.overcast.Cwqcn6hB_2dwXh4.jpg)![](/_astro/style.watercolor.DPW-cvt6_ZBhvN8.jpg)Snow Watercolor Show on stage Remove 1 of 6 parts Copy ## Image Prompt Structure: The Six Parts of a Prompt A prompt is a subject and five choices about it. The generator above writes them in this order. Subject What the picture is of. Start here; every other part describes it. Action What the subject is doing, or how it is turned. Setting Where it is. Camera How the camera sees it: the angle, the distance, the lens. Light Where the light comes from and what color it is. Light sets the mood. The generator's choices: window light, golden hour, overcast, backlit, candlelight, neon, studio softbox, moonlight. Style The finish: a photo, a film still, a painting. ### The Fox Prompt, Word by Word 1. Subjecta fox in a misty forest 2. Actiona fox in a misty forest, caught in a gust of wind 3. Settinga fox in a misty forest, caught in a gust of wind, in fresh snow 4. Cameraa fox in a misty forest, caught in a gust of wind, in fresh snow, shot from a low angle 5. Lighta fox in a misty forest, caught in a gust of wind, in fresh snow, shot from a low angle, golden hour sunlight 6. Stylea fox in a misty forest, caught in a gust of wind, in fresh snow, shot from a low angle, golden hour sunlight, 35mm film photo, natural grain ### The Portrait Prompt, Word by Word 1. Subjectphotorealistic portrait of a blonde woman with blue eyes 2. Actionphotorealistic portrait of a blonde woman with blue eyes, seen in profile 3. Settingphotorealistic portrait of a blonde woman with blue eyes, seen in profile, in an old library 4. Cameraphotorealistic portrait of a blonde woman with blue eyes, seen in profile, in an old library, shallow depth of field, blurred background 5. Lightphotorealistic portrait of a blonde woman with blue eyes, seen in profile, in an old library, shallow depth of field, blurred background, soft window light from the side 6. Stylephotorealistic portrait of a blonde woman with blue eyes, seen in profile, in an old library, shallow depth of field, blurred background, soft window light from the side, cinematic film still, muted color grade ### The Still Life Prompt, Word by Word 1. Subjecta ceramic teapot on a wooden table 2. Actiona ceramic teapot on a wooden table, with steam rising into the air 3. Settinga ceramic teapot on a wooden table, with steam rising into the air, in a cozy kitchen 4. Cameraa ceramic teapot on a wooden table, with steam rising into the air, in a cozy kitchen, shot from directly above 5. Lighta ceramic teapot on a wooden table, with steam rising into the air, in a cozy kitchen, shot from directly above, warm candlelight 6. Stylea ceramic teapot on a wooden table, with steam rising into the air, in a cozy kitchen, shot from directly above, warm candlelight, oil painting, visible brushstrokes ## AI Image Prompt Examples The recipes above, written out in full for the fox. Swap in your own subject and keep the rest. - Golden Hour Film a fox in a misty forest, caught in the wind, mountain meadow, telephoto lens, golden hour, 35mm film - Neon Street a fox in a misty forest, in motion, city street, wide-angle lens, neon, cinematic still - Quiet Study a fox in a misty forest, at rest, old library, shallow focus, window light, oil painting - Snow Watercolor a fox in a misty forest, partly hidden, fresh snow, close-up, overcast, watercolor ### From the Prompt Gallery Every image in the Private Diffusion prompt gallery keeps the exact prompt, model, and seed behind it. - [Interiors AI Image Prompts](/prompts/interiors) - [People AI Image Prompts](/prompts/people) - [Nature AI Image Prompts](/prompts/nature) - [Cinematic AI Image Prompts](/prompts/cinematic) - [Travel AI Image Prompts](/prompts/travel) - [Architecture AI Image Prompts](/prompts/architecture) - [Sci-Fi AI Image Prompts](/prompts/sci-fi) - [Editorial AI Image Prompts](/prompts/editorial) - [Skyborne pirate ship over misted hills](/prompts/skyborne-pirate-ship-over-misted-hills-z-image-turbo) - [Courier waits beneath floating lanterns](/prompts/courier-waits-beneath-floating-lanterns-anima) - [A staircase folding into clouds](/prompts/a-staircase-folding-into-clouds-kroma-turbo) [Browse every prompt](/prompts) ## How Do You Write a Prompt for Your Model? Every model reads up to 512 tokens, roughly 2,000 characters. On models with Rewrite Prompt, Studio turns it off once a prompt passes 200 characters, and you can turn it back on. Pick a model to see how it reads a prompt; the generator above writes in that form. ![](/model-logos/circlestone-labs.png)Anima![](/model-logos/black-forest-labs.svg)Flux.2 Klein 4BMFMage Flow Turbo![](/model-logos/krea.svg)Krea 2 Turbo![](/model-logos/lodestones.webp)Kroma Turbo![](/model-logos/tongyi-lab.png)Z-Image-Turbo![](/model-logos/rundiffusion.png)Juggernaut Z Fast![](/model-logos/baidu.webp)ERNIE Image Turbo![](/model-logos/lodestones.webp)Chroma1-Flash![](/model-logos/boogu.png)Boogu Image Turbo ![](/model-logos/circlestone-labs.png)Anima8 GB iPhone or iPad, any Mac Reads tags or sentences. No negative prompt; say what you want in the frame. 1. Write lowercase tags with spaces, not underscores, separated by commas.[Anima model card](https://huggingface.co/circlestone-labs/Anima) 2. Order the tags: count, character, series, artist, then everything else. Quality tags are optional; if you use them, they go first. An artist tag starts with @.[Anima model card](https://huggingface.co/circlestone-labs/Anima) 3. If you write sentences instead, write at least two: a very short prose prompt gives unexpected results.[Anima model card](https://huggingface.co/circlestone-labs/Anima) Example Prompt 1girl, solo, red fox ears, misty forest, soft morning light, looking at viewer, detailed background [Anima in Private Diffusion](/models/anima) ![](/model-logos/black-forest-labs.svg)Flux.2 Klein 4B8 GB iPhone or iPad, any Mac Reads sentences. Private Diffusion adds a negative prompt through Normalized Attention Guidance. Its system template uses part of the 512-token budget. 1. Describe the picture like a novelist, not as a keyword list.[Black Forest Labs prompting guide](https://docs.bfl.ai/guides/prompting_guide_flux2) 2. Put the subject first, then what it does, the style, and the context, in about 30 to 80 words.[Black Forest Labs prompting guide](https://docs.bfl.ai/guides/prompting_guide_flux2) 3. Put any text that should appear in the image inside quotation marks.[Black Forest Labs prompting guide](https://docs.bfl.ai/guides/prompting_guide_flux2) Example Prompt A red fox pauses on a mossy log in a misty pine forest at dawn. Low golden light rakes through the trees behind it, catching the fur along its back. Shot on a telephoto lens with shallow focus, in the style of a quiet nature documentary still. [Flux.2 Klein 4B in Private Diffusion](/models/flux-2-klein-4b) MFMage Flow Turbo12 GB iPhone or iPad, 16 GB Mac Reads sentences. No negative prompt; say what you want in the frame. 1. Describe the scene in full sentences and name the medium: photograph, oil painting, product render. Example Prompt An oil painting of a red fox at rest beside a candlelit window in an old study, warm flame light on one side of its face, deep shadow on the other, thick impasto brushwork. [Mage Flow Turbo in Private Diffusion](/models/mage-flow-turbo) ![](/model-logos/krea.svg)Krea 2 Turbo16 GB iPad or Mac Reads sentences. No negative prompt; say what you want in the frame. 1. Write in natural language; long, detailed prompts give the best results.[Krea 2 prompting guide](https://github.com/krea-ai/krea-2/blob/main/docs/prompting.md) 2. Quote any text that should render in the image.[Krea 2 prompting guide](https://github.com/krea-ai/krea-2/blob/main/docs/prompting.md) Example Prompt A photograph of a red fox standing on a frost-covered meadow at golden hour, long shadows stretching toward the camera, warm backlight on its fur, mountains soft in the distance, 85mm lens, shallow depth of field, natural film grain. [Krea 2 Turbo in Private Diffusion](/models/krea-2-turbo) ![](/model-logos/lodestones.webp)Kroma Turbo16 GB iPad or Mac Reads sentences. Reads a negative prompt. 1. Write the prompt as you would for Krea 2 Turbo: natural language, the more specific the better. 2. Use the negative prompt for the usual suspects: extra limbs, text, watermarks, blur. Example Prompt A red fox curled at rest on a sunlit windowsill in an old library, dust in the light, shelves of worn books behind it, painted in loose oil strokes with visible canvas texture. [Kroma Turbo in Private Diffusion](/models/kroma-turbo) ![](/model-logos/tongyi-lab.png)Z-Image-Turbo12 GB iPhone, 8 GB iPad, any Mac Reads sentences. Private Diffusion adds a negative prompt through Normalized Attention Guidance. Its system template uses part of the 512-token budget. 1. Keep the subject, its count, its action, and its colors exact; the model follows what you name.[Z-Image prompt enhancer](https://github.com/Tongyi-MAI/Z-Image) 2. Put on-image text in double quotes and say its font, layout, and position.[Z-Image prompt enhancer](https://github.com/Tongyi-MAI/Z-Image) 3. Stay objective and concrete; skip "8K", "masterpiece", and other quality words.[Z-Image prompt enhancer](https://github.com/Tongyi-MAI/Z-Image) Example Prompt One red fox walking through fresh snow in a birch forest, seen from a low angle, overcast daylight, pale blue shadows, a wooden sign in the foreground reading "TRAILHEAD" in white serif capitals. [Z-Image-Turbo in Private Diffusion](/models/z-image-turbo) ![](/model-logos/rundiffusion.png)Juggernaut Z Fast12 GB iPhone, 8 GB iPad, any Mac Reads sentences. Private Diffusion adds a negative prompt through Normalized Attention Guidance. Its system template uses part of the 512-token budget. 1. Write natural descriptive sentences, not tag lists.[RunDiffusion Juggernaut Z prompt guide](https://www.rundiffusion.com/juggernaut-z-prompt-guide) 2. Put any exact on-image text first, then the subject, its attributes, the action, the environment, the lighting, and the composition.[RunDiffusion Juggernaut Z prompt guide](https://www.rundiffusion.com/juggernaut-z-prompt-guide) 3. Name colors; the model renders the ones you write.[RunDiffusion Juggernaut Z prompt guide](https://www.rundiffusion.com/juggernaut-z-prompt-guide) Example Prompt A photograph of a red fox in profile on a quiet city street at night, neon shop signs in magenta and cyan reflecting on wet asphalt, wide-angle lens, cinematic framing, shot on digital with a slight cool grade. [Juggernaut Z Fast in Private Diffusion](/models/juggernaut-z-fast) ![](/model-logos/baidu.webp)ERNIE Image Turbo16 GB iPad or Mac Reads sentences. No negative prompt; say what you want in the frame. 1. Write the whole brief as one long paragraph: the camera angle, where each element sits, the light direction, and the palette.[ERNIE-Image model card](https://huggingface.co/baidu/ERNIE-Image) 2. It is strong on dense text and multi-object layouts, so spell out exact words and positions.[ERNIE-Image model card](https://huggingface.co/baidu/ERNIE-Image) Example Prompt Overhead view of a wooden table in a cozy kitchen: a ceramic teapot at the center with steam rising, two cups to its left, a folded linen napkin to its right, morning window light from the top of the frame, warm cream and terracotta palette, a handwritten card reading "tea time" at the bottom edge. [ERNIE Image Turbo in Private Diffusion](/models/ernie-image-turbo) ![](/model-logos/lodestones.webp)Chroma1-Flash16 GB iPad or Mac Reads sentences. Private Diffusion adds a negative prompt through Normalized Attention Guidance. 1. Natural language works better than tag lists.[Chroma1-HD model card (third party)](https://huggingface.co/lodestones/Chroma1-HD) 2. Avoid very short prompts; give the model a full sentence or two to work from.[Chroma on Civitai (third party)](https://civitai.com/models/1330309/chroma) Example Prompt A watercolor of a red fox half hidden behind a snow-covered fir branch, soft overcast light, loose wet-on-wet washes of gray and rust, white paper showing through the highlights. [Chroma1-Flash in Private Diffusion](/models/chroma1-flash) ![](/model-logos/boogu.png)Boogu Image Turbo16 GB iPad or Mac Reads sentences. No negative prompt; say what you want in the frame. Its system template uses part of the 512-token budget. 1. Write long descriptive paragraphs; the model rewards detail. 2. Small fonts and dense text drift, so keep any on-image text short and large.[Boogu Image README](https://github.com/boogu-project/Boogu-Image) 3. Frame faces large, since small ones lose detail, and avoid crowded or unusual poses where hands are visible.[Boogu Image README](https://github.com/boogu-project/Boogu-Image) Example Prompt A red fox standing in a mountain meadow of tall grass, wind pushing the grass and its fur to the left, late afternoon sun low behind it, a wide landscape with soft peaks on the horizon, rendered as a clean 3D illustration with gentle global illumination. [Boogu Image Turbo in Private Diffusion](/models/boogu-image-turbo) ## Questions About AI Image Prompts - ### What Are Good AI Image Prompt Examples? The recipes under the generator are complete prompts: pick one, change the subject, and the sentence updates. The Image Prompt Structure section builds each example subject up one part at a time, from the bare subject to the finished prompt. Every image in the Private Diffusion prompt gallery keeps the exact prompt, model, and seed that made it, so you can copy any of them. [Browse every prompt](/prompts) - ### What Is a Negative Prompt? A negative prompt lists what you do not want in the picture, such as "blurry, text, watermark". It is a separate field, never part of the main prompt. In Private Diffusion, Flux.2 Klein 4B, Kroma Turbo, Z-Image-Turbo, Juggernaut Z Fast, and Chroma1-Flash read it. The other models ignore it, which is why the For Studio panel hides the field when you pick one of them. - ### How Do I Write a Prompt for an AI Image? Start with a subject you can picture, then add one part at a time: what it is doing, where it is, how the camera sees it, the light, and the style. Each part changes the image in a way you can see, and the stage above shows that change one word at a time on the same seed. Most models read plain sentences; Anima reads comma-separated tags. - ### How Do I Prompt Flux? Flux.2 Klein 4B reads plain sentences. Describe the picture the way you would describe a photograph to a friend: the subject first, then what it is doing, the style, and the context (where it is and how it is lit), in about 30 to 80 words. Put any text that should appear in the image in quotation marks. Pick Flux.2 Klein 4B in the Writing for row and the generator composes in that form. [Flux.2 Klein 4B in Private Diffusion](/models/flux-2-klein-4b) - ### Will the Prompt Work in Midjourney or Other Generators? Yes. The prompt is plain text with nothing specific to Private Diffusion in it, so you can paste it into Midjourney, Nano Banana, or any other image generator. Models read prompts differently, so the Writing for row rewords yours for the model you pick, and the model guide shows what each one reads best.