Drop any photo — get the recipe back as a clean text prompt. HotPrompt's free image-to-prompt AI reads the picture through a senior photographer's and art director's lens, then writes the prompt you'd need to recreate the shot in Midjourney, Flux, Stable Diffusion (Civitai / ComfyUI), DALL·E, ChatGPT, Perchance, ImageFX, Leonardo, Clipdrop or Picsart. The best way to write image prompts is to start from a reference — this is that starting point.
Before vs After
Most image-to-prompt converters just describe what's in the picture. HotPrompt reads it the way a photographer would — the lens, the lighting, the grade, the composition rules — then writes the prompt a senior art director would brief.
“A 1080×1080 JPEG of a moody cyberpunk café at dawn.”
# Subject & composition A weathered barista mid-pour at a corner counter; viewer placed slightly below eye line; rule-of-thirds with subject on right intersect; deep alley vanishing point on left frame; foreground cup catches a sliver of neon as a leading line. # Camera, lens & shot Mirrorless full-frame; 35mm prime; f/1.8 shallow depth of field; ~1/60s handheld; subtle motion blur on pouring steam. # Lighting Cool blue key from neon signage outside frame-right; warm tungsten fill from interior pendant; soft rim on the barista's shoulder from window backlight; mixed color temperature — cool exterior, warm interior. # Color & grade Palette: teal, magenta, amber, deep navy. High contrast, lifted shadows. Cinestill 800T grade with mild halation on highlights. Roger Deakins teal-orange tilt. # Style & era Editorial cyberpunk realism; mid-2020s aesthetic; nods to Wong Kar-wai jewel tones and Saul Leiter color-block compositions. # Texture & detail Visible film grain at ISO 1600-equivalent; skin pores readable on the barista; condensation on the cup; wet pavement reflections; specular highlight on the espresso surface. # Prompt for Flux A weathered barista mid-pour at a corner counter of a moody cyberpunk café at dawn, 35mm prime, f/1.8 shallow depth of field, cool blue neon key from frame-right mixed with warm tungsten fill, Cinestill 800T halation, teal and magenta palette with deep navy shadows, lifted contrast, Saul-Leiter color-block composition with rule-of-thirds placement, visible film grain, photorealistic editorial cyberpunk. # Avoid warped fingers, melting cup, text on signs, low resolution, oversaturated neon, plastic skin
How it works
Four steps from a photo on your desktop to a prompt you can paste straight into Midjourney, Flux, Civitai or ComfyUI — under thirty seconds.
Drag in any PNG, JPEG or WEBP up to 20 MB. Portrait, landscape or square — the optimizer reads aspect ratio and adapts the output.
Subject, composition, lens, aperture, lighting, color, palette, grade, film stock reference, era — every dimension a senior DP would name.
Midjourney, Flux, Civitai, ComfyUI, DALL·E (ChatGPT), Perchance, ImageFX, Leonardo, Picsart, Clipdrop — output reshapes for each.
Output is plain text. Paste into your favorite AI image tool, render once, then tweak the prompt to remix the look.
Every output reads like a photographer briefing a re-shoot — and lands in the exact format your target AI image tool wants.
Subject + composition (rule of thirds, leading lines, negative space, golden ratio).
Camera, lens, aperture and shutter inferred from depth of field and motion clues.
Lighting setup (key / fill / rim, hard vs soft, time of day, color temperature).
Palette + grade with a named film stock or photographer reference (Portra, Cinestill, Deakins, Saul Leiter).
Tool-aware output — flat tags for Midjourney, weighted tags for Civitai/ComfyUI, prose for Flux and DALL·E.
Negative-prompt 'Avoid' line that kills the usual AI artifacts (warped fingers, plastic skin, text glitches).
Pick the model you're submitting the prompt to. The director's analysis stays the same, but the final 'Prompt for X' section is shaped for that tool's input.
Flat, comma-delimited tag string ending with --ar, --stylize, --v 6 and a --no negative line.
Subject-first descriptive prose, no parameter flags — exactly how Flux models prefer to be briefed.
Tag list with parenthesis weights like (subject:1.2), separate negative prompt block — the format SD samplers parse natively.
Structured positive + negative blocks ready to paste into CLIPTextEncode nodes, with weight hints where the look depends on emphasis.
Natural-language paragraph — DALL·E 3 grades better on prose than tag-soup.
Tag list with brief style anchors — tuned for Leonardo's PhotoReal and Ideogram's typography strengths.
Midjourney-style flat tag prompts. Perchance image to prompt, ImageFX image to prompt and Pixverse image to prompt all read the same shape.
Compact tag list — Picsart image to prompt and Clipdrop image to prompt prefer shorter, punchy briefs.
Most image-to-prompt converters stop at 'what's in the picture'. We go six layers deeper — these are the dimensions a senior photographer would brief.
What's in frame, where it sits — rule of thirds, leading lines, negative space, frame-within-a-frame, golden ratio. The skeleton of every great shot.
DSLR vs mirrorless vs phone vs cinema vs medium format. Lens guess (24mm wide / 35mm / 50mm / 85mm portrait / macro / tilt-shift). Aperture inferred from depth of field.
Key light direction, fill, rim, hard vs soft quality. Time of day (golden hour, blue hour, overcast, studio strobe). Color temperature — warm, cool or mixed.
Dominant palette named (3-5 colors). Contrast (low-key, high-key, mid-tone). Film stock reference — Kodak Portra 400, Fuji Pro 400H, Cinestill 800T, Wong Kar-wai jewel tones.
Editorial fashion, documentary, street, portrait, landscape, still life, product, cyberpunk, fantasy, film noir. Photographer reference baked in.
Skin pores, fabric weave, surface roughness, film grain, digital noise, specular highlights. The difference between 'AI flat' and 'photorealistic'.
Once you've seen the recipe, you know how to brief one yourself. Five rules — the optimizer follows them by default; useful to know when you're tweaking the result.
The first five words decide composition. 'A weathered barista mid-pour' beats 'a coffee shop scene' every time.
35mm prime, 50mm portrait, 85mm telephoto, 24mm wide, 100mm macro. Lens is the difference between amateur snap and pro shot.
Key light direction, fill, rim. Time of day. The lighting setup is what gives the image mood — AI fills it with stock daylight if you don't.
Film stock (Cinestill, Portra, Provia) or photographer reference (Saul Leiter, Roger Deakins, Helmut Newton). One name snaps the look out of stock.
Warped fingers, plastic skin, garbled text, low resolution. The negative-prompt line kills the usual AI artifacts before they render.
Anywhere a reference image needs to become a prompt — to recreate it, to brief an AI tool with it, or to teach yourself why it works.
Hand over a reference photo; get the prompt your team uses to render brand-aligned variations across Flux, Midjourney and DALL·E.
Pin a hero shot from Instagram, Pinterest or a competitor's site; HotPrompt tells you exactly how it was lit, lensed and graded.
A Civitai image-to-prompt and ComfyUI-ready output with weighted tags and separate negatives — ready to wire into your node graph.
Drop a finished shot and get the brief you'd hand to an assistant — useful for client refs, education or your own portfolio notes.
Image to prompt extension territory: pixel-peep a Leonardo AI image to prompt, copy the recipe, render variations in Flux or DALL·E.
Pinterest moodboards become images-to-prompt creative writing seeds — generated descriptions feed straight into screenplay and novel prep.
Questions
Start from a reference image. Even great prompt engineers struggle to brief lighting and grade from scratch — looking at a shot you love and naming what makes it work is faster and more accurate. This tool is that starting point: drop the photo, get the recipe back.
Yes — Civitai image-to-prompt and ComfyUI workflows are first-class targets. Pick either in the target dropdown and you get a tag list with parenthesis weights (like (subject:1.2)) and a separate negative block, ready to paste into CLIPTextEncode nodes or the Civitai prompt field.
All three. Flux gets descriptive prose (it grades better that way), SD/Civitai gets weighted tags + negatives, and Midjourney gets a flat tag string ending in --ar, --stylize and --no. The director's analysis is the same; only the final 'Prompt for X' section reshapes.
Most image to prompt converter and image-to-prompt extension tools just caption the picture ('a café with neon signs'). We read it like a photographer — six dimensions including lens, lighting setup, grade reference and composition rules — then write the brief in your target tool's format.
Yes — the generated image prompt is the cleanest possible starting point for any image to video generator free workflow (Kling AI image-to-video, Pixverse image-to-video, Runway, Hailuo). Take HotPrompt's output, add a motion line ('subject walks toward camera, gimbal dolly') and feed both into your image-to-video tool.
Sign-up is free and takes 10 seconds (no credit card, no email verification). Every account gets 10 free credits per day and an image analysis costs 10 credits, so the free tier covers one analysis daily forever. No paywall, no hidden 'pro' gating on the result.
Yes. Pick 'DALL·E (ChatGPT)' as the target and the output is natural-language prose — DALL·E 3 grades better on prose than tag-soup. Paste it into ChatGPT's image tab and render.
For that workflow, use the regular Optimize page with category set to Image. This tool is purpose-built for upload-first 'photo to image prompt' and 'describe image to prompt' workflows where you have a reference you want to reverse-engineer.
More tools
Five birthday card concepts — AI-image prompt + matching greeting message, tailored to your recipient.
Upload any image — get the prompt that would have generated it. Photographer-grade analysis.
Build a custom zsh PROMPT string with live preview — no LLM, no credits, instant.