Yes, for a mood-level preview. Image tools such as ChatGPT, Gemini and Midjourney can show a customer the colour palette, shape, rough size and feel of an arrangement in minutes, even placed in a photo of their venue. They can't be trusted on exact varieties, stem counts, mechanics or season, so present a mock-up as the feel, not a promise.
The bigger risk is a beautiful image you can't build. Image models happily produce cascades held up by nothing, flowers in colours they don't grow in, and peonies in October. A customer who falls in love with that picture will be disappointed by your real, excellent arrangement. So the most useful mock-ups start from your own work, or are prompted in florist's terms, and are always shown with a clear note of what will differ.
What a mock-up shows reliably, and what it gets wrong
| Usually reliable | Often wrong |
|---|---|
| Overall colour palette and how colours sit together | Exact flower varieties; similar flowers merge into one invented flower |
| Shape: dome, loose and garden-style, cascade, low and long | Stem counts and density; it adds or removes flowers at will |
| Rough scale against a vase, table or arch in a photo | Precise dimensions; "40 cm high" rarely comes out at 40 cm |
| Mood: wild and seasonal versus formal and structured | Mechanics: foam-free structures, weight, how an arch is actually fixed |
| How a piece might look in a venue's light and setting | Seasonality and availability; it doesn't know what your wholesaler has |
Anything in the right-hand column is your job to explain. The left-hand column is where mock-ups genuinely help customers who find it hard to picture a design from words and a mood board.
Three ways to make a mock-up, from quickest to most accurate
1. Describe it from scratch
Text-to-image is quickest: type a description, get four options. ChatGPT and Gemini both generate images inside the chat, and Midjourney is a dedicated image tool with plans from about $10 a month. The results look polished but drift furthest from what you'd build, because the model invents freely. Good for early mood conversations ("more like this, or more like that?"), weak for anything the customer might treat as a specification.
2. Add an arrangement to a photo of the venue or table
Upload the customer's photo of their table, mantelpiece or ceremony space and ask the tool to add an arrangement. Gemini's image editing (Google calls its current image model Nano Banana 2) and ChatGPT's image editing can both work from an uploaded photo. This is where mock-ups earn their keep, because scale and setting matter more to the customer than exact varieties. Check the proportions carefully: tools often shrink or enlarge the room to fit the flowers.
A quick sum catches most scale errors. Ask the venue for the size of its tables; say its rounds are 180 cm across. A 60 cm trough should then span a third of the table's width, and a 25 cm height should sit at roughly the height of a tall wine glass. If the trough in the image stretches most of the way across the cloth, or towers over the glassware, the tool has scaled it to look impressive rather than to match your quote. Regenerate with the table size in the prompt ("the table is 180 cm across; the arrangement covers one third of its width"), or show the customer the image with the real dimensions written beside it.
3. Edit a photo of your own past arrangement
The most accurate route, and the one I'd use for anything important: take a photo of a piece you've actually made with a similar structure, and ask the tool to change the colour palette or swap one element. "Change the roses in this arrangement to a deep burgundy and replace the white stocks with cream" keeps your mechanics, proportions and style, because it starts from something real. The customer sees your work, recoloured, rather than an AI's idea of floristry.
Edits drift unless you pin down what must stay the same. An illustrative first try with the prompt above came back with the roses a convincing burgundy, but the model had also darkened the eucalyptus to near-black, swapped the ceramic jug for a glass vase and cropped the image tighter. The customer would be approving a different piece. Adding one sentence fixes most of it: "Change only the colours of the roses and stocks. Keep the container, foliage, shape, background, lighting and framing exactly as in the photo." If one element still changes, ask for that element back on its own ("put the original white ceramic jug back") rather than starting again. In Midjourney, whose default model has been V8.2 since July 2026, this kind of change goes through its Edit Model, which replaced the older Omni Reference and Retexture features.
One edge case comes up often: a customer sends a photo of another florist's arrangement and asks for "exactly this". Don't upload that photo and edit it into your proposal; it is someone else's work, and the result would still be their design with your name on it. Describe what the customer likes about it (the palette, the height, the trailing foliage), build a mock-up from one of your own pieces with those qualities, and show the two side by side so the customer can see what you'd bring to it.
Reading the customer's inspiration photos first
Before making anything, AI can help you read what the customer has sent. Upload their three favourite inspiration images and ask:
These are a customer's inspiration photos for wedding table flowers.
For each: likely flowers and foliage (say how confident you are),
the colour palette in plain words, the shape and approximate height,
and which flowers might be hard to source in [month].
Then summarise what the three have in common.An illustrative answer: "All three are low and loose, in blush, cream and dusty pink with lots of trailing greenery. Photo 1: garden roses (high confidence), sweet peas (medium), jasmine vine. Photo 2: roses and lisianthus (medium). Photo 3: peonies (high), which may be hard to source in October." Useful, and mostly right, but check it: the "roses" in photo 2 are actually lisianthus throughout, a common confusion, and the sweet peas in photo 1 are clematis. The summary of what the photos share is the valuable part. It gives you a sentence to confirm with the customer ("so, low, loose and romantic, in blush and cream, with trailing greenery?") before you design anything, and it flags the peony problem early.
A prompt that gets closer to something you could build
Image tools respond better to the vocabulary florists already use. A general-purpose template:
Photorealistic photo of a [shape: low, long, loose garden-style]
flower arrangement, about [size] wide and [size] high, in a
[container], on a [surface] in [setting and light].
Flowers: [list only real varieties you'll use, with colours].
Foliage: [list]. Colour palette: [colours].
Style: [natural and seasonal / structured and formal].
No flowers other than those listed. No floating stems. Realistic
proportions. Natural daylight, no dramatic filters.Filled in for an autumn wedding centrepiece:
Photorealistic photo of a low, long, loose garden-style flower
arrangement, about 60 cm wide and 25 cm high, in a low rustic
wooden trough, on a round banquet table with white linen, in a
barn with warm evening light. Flowers: rust and peach garden
roses, cream spray roses, terracotta dahlias, dried bunny tails.
Foliage: eucalyptus, trailing ivy. Colour palette: rust, peach,
cream, sage. No flowers other than those listed. No floating
stems. Realistic proportions. Natural light, no filters.Illustrative result, and what you'd fix: the image shows a lovely low trough with the right palette, but the model has added three orange ranunculus that weren't listed, made the dahlias twice the size of the roses, and trailed the ivy halfway across the table. The palette and shape are right, so it's good enough to confirm the direction. Tell the customer: "The colours and shape are what we're agreeing; the exact flowers and size will be as in my quote." Regenerating to get every stem right wastes time, because it rarely converges.
Centrepieces for 12 barn tables, from photo to quote
Consider a customer planning a 90-guest autumn wedding at a barn venue, with 12 round tables. They've sent three inspiration photos and the venue's own photo of the room. They like "autumn colours but not orange" and are worried about centrepieces blocking conversation.
- Before the consultation (10 minutes). The florist edits a photo of a low trough arrangement from a past wedding, asking the tool to recolour it in rust, blush and cream, and generates a second version with burgundy instead of rust.
- Placing it in the room (5 minutes). Using the venue photo, the florist asks for the trough arrangement on each visible table, at 25 cm high. The result shows the room full of centrepieces, which helps the customer see whether it looks too sparse or too busy.
- At the consultation. The customer picks the blush version, asks for the height to stay low, and adds candles. The florist explains that the dahlias in the image might be swapped for chrysanthemums depending on what's good that week.
- Quote. The quote lists the actual flowers per centrepiece, the trough, candles and set-up, and includes the mock-up image with a line saying it is an AI-generated illustration of palette and shape.
Fifteen minutes of preparation replaced a lot of describing, and the customer's worry about height was answered with a picture rather than a promise. Costs in this case: nothing beyond the florist's existing paid chat assistant at about $20 a month. For writing the proposal that accompanies the mock-up, see writing wedding flower proposals with AI.
Contract work uses the same method with a different customer. A florist pitching weekly reception-desk flowers to a small office, say, can photograph the desk on a site visit and mock up three seasonal options in the space: a spring piece in white and green, a summer piece in the company's blue-and-yellow tones, an autumn piece in bronze. The office manager can compare them against the desk, the logo wall and the lighting in a way a price list never shows. One thing to say plainly with this kind of mock-up: flowers come in tones, not exact brand colours. An image tool will happily produce a delphinium in the precise blue of the company logo; the real stems will be close, and will vary week to week.
Showing it to a customer without over-promising
Put a short note under every mock-up, in the email and on any proposal page. A filled-in example:
This image is an AI-generated illustration of the colour palette, shape and approximate size we discussed. Your arrangement will be made by hand from the flowers listed in your quote, chosen for quality in the week of your wedding, so varieties, stem counts and exact proportions will differ. If any flower isn't at its best that week, we'll substitute something in the same colour and style and let you know.
Say the same thing out loud at the consultation. The written note protects you if there's a dispute; the conversation is what actually sets expectations. And never send a mock-up for a piece whose key flower is out of season on the date. Spotting that is your expertise, not the model's.
To check whether your mock-ups are setting expectations well, keep the mock-up and a photo of the finished piece side by side in each job's folder. After a season, look through the pairs with one question: where did customers' comments on the day differ from what the image suggested? If several said something like "I thought it would be fuller", your images are denser than your work; add "airy, with space between stems" to your prompt and a line to the note. If nobody noticed any difference, the method is doing its job.
Four realistic ways mock-ups go wrong
- The colour that doesn't exist. A customer approves a mock-up with vivid electric-blue roses. There's no natural rose in that colour; dyed stems look nothing like the image. The florist should have spotted it before sending, and now has an awkward conversation. Check every flower in the image against what can really be sourced in that colour.
- The height nobody measured. A mock-up of a tall arrangement on a top table looked elegant, but in the room it blocked the couple from the guests' view. The image showed a plausible height; it didn't show sightlines from seated guests. For anything taller than about 30 cm on a dining table, check sightlines at the venue or from its floor plan.
- The confident caption. Asked to "write a short description to go with this mock-up for the customer", an assistant produced (illustrative): "Your stunning centrepieces will feature lush garden roses, velvety dahlias and cascading ivy, exactly as pictured." Two words undo the careful note under the image: "exactly as pictured". Tell the assistant to describe the palette and shape only, and to say the image is an illustration, or write the two sentences yourself.
- The image that travelled. A customer posted the AI mock-up online, tagging the florist, as "our wedding flowers". Followers then expected that image from the shop. Labelling the mock-up clearly, and sending a photo of the real arrangement after the wedding, limits the damage. How to label AI images is covered in labelling AI-generated images on social media.
Costs, privacy and watermarks
- Cost. ChatGPT Plus and Google's AI Pro plan both include image generation, within limits that change often. Free tiers allow a small number of images. Midjourney starts at about $10 a month for its entry plan, with no free trial.
- Privacy. On Midjourney's Basic and Standard plans, your images can appear on its public Explore page; its Stealth Mode, which keeps creations private on the website, is only on the Pro and Mega plans. Think twice before uploading a customer's venue photo or home interior to any tool, and never upload photos with identifiable people without permission.
- Watermarks. Gemini marks the images it creates with an invisible SynthID watermark and a visible one. Leave visible watermarks in place on mock-ups; they help make clear the image isn't a photo of your work.
Whether AI images can be used commercially at all is a separate question, covered in whether you can legally use AI-generated images in your marketing. For mock-ups kept inside a consultation and clearly labelled, the practical risks are small. The bigger discipline is keeping them out of your portfolio. For the rest of the shop's day-to-day AI use, see how florists use AI for orders, reviews and social posts.
AI flower mock-ups: more questions
Can I use AI mock-ups to advertise my arrangements?
Not as if they were your work. Customers who order from an image expect the flowers in it, and an AI image of something you've never made misleads them. Use AI mock-ups in consultations, clearly labelled, and advertise with photos of arrangements you actually made. If you ever post an AI image, label it as AI-generated.
Can AI identify the flowers in a customer's inspiration photo?
Often, but not reliably. General assistants can suggest likely varieties from a photo, and they're right on common flowers much of the time, but they confuse similar ones such as ranunculus and garden roses, or lisianthus and roses. Use the suggestion as a starting point for the conversation, then identify the stems yourself.
Is it worth doing for everyday bouquets?
Rarely. For a $50 hand-tied bouquet, a photo of a similar past bouquet and a colour swatch are quicker and more honest. Mock-ups pay off for larger or unusual commissions where the customer struggles to picture the result: wedding and event pieces, installations, corporate displays and repeat contract work.
Which tool should a florist start with?
Start with whichever general assistant you already use, since ChatGPT and Gemini both generate and edit images, including edits to a photo you upload. Try editing a photo of your own past arrangement first, because it keeps the result closest to what you can build. Consider a dedicated image tool only if you make mock-ups every week.
Further reads
- How Florists Can Prepare for Valentine's and Mother's Day With AI — Plan stock and staff for the busiest florist days.
- AI Model Shots for Boutiques: Try-On Images Without a Photoshoot — The same honesty question for fashion try-on images.
- AI Room Visualisation From Client Photos: A Designer's Workflow — A designer's workflow for editing client photos with AI.
- Brand Guidelines for AI Images: Colours, Style and Limits — Set rules for colours and style in any AI images.
- Can ChatGPT Read PDFs, Spreadsheets and Photos? What Breaks — What ChatGPT can and can't read from a photo.
- Is Midjourney Worth It for Small Business Marketing Images? — Whether a paid image tool earns its fee.
- AI Garden Design Visuals in Quotes: Do They Help Close Jobs? — When an AI-edited photo of the client's own garden helps a landscaping quote get signed, when it backfires, and a fair way to test it.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Google's Gemini image generation overview page (Nano Banana 2, SynthID and visible watermarks); Midjourney documentation on plans, Stealth Mode, the V8.2 default model and the Edit Model.