Product Description From Image: Listing Copy From a Product Photo

A product description from image is listing copy you write from a product photo: a title, feature bullets, and a short SEO blurb for a PDP or marketplace page. You attach the still, name the channel and brand voice, and demand copy that names visible features plus catalog specs you paste. You leave with ROLE / TASK / FORMAT product description prompts for GPT-5.6 Sol, Claude Fable 5, Claude Opus 5, and Gemini 3.5 Flash, listing kits, a photo-then-copy workflow, a CMS paste path, and a mistakes list. Product photography prompts on this blog write generator language for new packshots. Stay here for copy. Soft scaffold: https://promptmake.net/text

What a product description from image is for

You search "product description from image" when a packshot or lifestyle still must become words a shopper reads and a crawler indexes. The finish line is listing copy: a title, feature bullets, and a short blurb you paste into Shopify, Amazon, Etsy, WooCommerce, or a custom product detail page (PDP). A vision chat can see glaze, a zipper, a heel height. It cannot see capacity, RFID, battery hours, or your target query unless you paste those facts. Your job is to attach one photo, feed catalog specs, and lock format so the model writes a listing instead of a caption.

An AI path does that job at catalog scale. You attach a cutout on white, an on-body apparel shot, or a three-quarter hero. You add CHANNEL (Amazon, Shopify, Etsy), BRAND_VOICE (spare, warm, technical), and CATALOG_SPECS the pixels hide. The model drafts a title, five bullets, and a 2 to 4 sentence blurb. You edit one pass for invented materials, fake measurements, and stuffed queries. You paste into the listing fields. That is the finish line. Midjourney flags, FLUX photography clauses, and SDXL negatives belong in the image-to-prompt guides on this blog.

This page supports the PromptMake product description use case at https://promptmake.net/use-case/product-description. The photo-in path there can return copy from an upload. Stay here for product description prompts you run in GPT-5.6 Sol, Claude Fable 5, Claude Opus 5, Claude Sonnet 5, Gemini 3.5 Flash, or Gemini 3.1 Pro: how to ask for title, bullets, and blurb a merchandiser can trust. Soft scaffold if the ask is messy: https://promptmake.net/text

Strong fit for:

• Shop teams who must turn a drop of SKU photos into titles and bullets without writing each listing from a blank field
• Marketplace sellers who need Amazon-style or Etsy-style copy that matches the photo plus a spec sheet

Skip this page if you want a generator prompt for a new packshot. Open the product photography and image-to-prompt posts for that job. Listing copy serves shoppers first. Crawlers index honest nouns in the title and the blurb.

How to write product description prompts from a photo

Vision chat models such as GPT-5.6 Sol, Claude Fable 5, and Gemini 3.5 Flash can see an attached product photo. They write the wrong kind of text if you ask them to describe the image. A caption names the mug and the marble. A listing needs a title a shopper can scan, bullets that tie a visible feature to a use, and a short blurb a crawler can index. Name the job in sentence one of the prompt: product listing copy for CHANNEL, in BRAND_VOICE, from the attached photo plus CATALOG_SPECS. The model sees glaze and a lid. It does not see 350 ml, dishwasher rules, or your keyword. Put those facts in the prompt. Use the ROLE / TASK / FORMAT skeleton next, then the spec checklist.

GPT-5.5 Instant is enough for a first pass on a quiet packshot on seamless white. GPT-5.6 Sol and Claude Opus 5 hold up better on busy lifestyle frames with props, on-body apparel, and photos that include a label you must quote. Gemini 3.5 Flash is a fast batch partner if you keep the same skeleton and swap IMAGE_CONTEXT. Name the model in your notes so a teammate can rerun the same ask. Vague "describe this product" chats produce paragraphs. Listing fields want a title, bullets, and a blurb you can defend against the pixels and the spec sheet.

Role, task, and format for listing copy

Copy this skeleton. Fill the brackets. Attach the photo in the same turn.

ROLE: You write e-commerce listing copy. You name visible features from the photo. You use CATALOG_SPECS for facts the pixels hide. You do not invent materials, measurements, certifications, or offers. You do not write Midjourney or FLUX prompts.

TASK: From the attached product photo, write listing copy for CHANNEL in BRAND_VOICE. Use IMAGE_CONTEXT for the product name and the query you want in the title. Use CATALOG_SPECS for capacity, size, compatibility, care, and claims you verified.

FORMAT: Return three labeled blocks. TITLE: one line, under TITLE_MAX characters. BULLETS: five lines, each one feature plus one buyer use, no leading emoji. BLURB: two to four sentences for the PDP. Then a one-line META under 160 characters if CHANNEL is a web store.

CHANNEL: [Amazon | Shopify | Etsy | WooCommerce | custom PDP]

BRAND_VOICE: [spare catalog | warm maker | technical spec]

IMAGE_CONTEXT: [SKU name, colorway, hero view: front / side / on-body / in-use]

CATALOG_SPECS: [capacity, dimensions, materials you verified, care, compatibility, warranty. Write unknown if a fact is missing.]

TITLE_MAX: [Amazon about 200 characters | Shopify nav 70 to 80 | Etsy 140]

RULES: Do not claim a sale, a rank, or a certification the photo and CATALOG_SPECS omit. Do not start TITLE with buy or cheap. If a printed word on the product matters (a model number on a label), include that word. If you cannot see a claimed material, omit it or mark unknown.

REMINDER: Shoppers read this next to the photo. Search engines store the title and the blurb. Keep both true.

That reminder keeps the model from writing ad slogans. Models love tidy boasts. Your rule forces a listing a merchandiser can ship. After the first reply, run a 30-second edit: delete a leather grain the photo does not show, drop a milliliter count you did not supply, cut a second title clause that restates the color.

Specs the photo cannot show

Pixels omit the facts marketplaces rank and shoppers compare. A matte black mug on marble can be 350 ml or 470 ml. A slim wallet can include RFID or skip it. Headphones hide battery hours. A knife hides steel grade and NSF status. Paste CATALOG_SPECS with each still. Skip them and the model guesses a number that will fail a returns ticket.

Add locale and audience when they change words. A US apparel page says sneakers. A UK page may say trainers if that is the catalog term. Match the store language so the title and the surrounding PDP agree. Add a do-not list for claims you will set in other fields: sale badges, "bestseller" ribbons, free-shipping banners. The photo shows a mug on marble. The sale lives in a CMS field. Keep them apart.

For apparel, paste size range, fabric composition, and care from the hangtag or the tech pack. The model can see a crew neck and a rib cuff. It cannot see 100 percent merino or a 30-degree wash. For electronics, paste battery, ports, and compatibility from the spec sheet. For food, paste net weight, allergens, and storage from the label you verified. IMAGE_CONTEXT names the SKU. CATALOG_SPECS is the source of truth for numbers.

Step-by-step workflow from photo to listing copy

A repeatable loop beats a lucky ChatGPT paragraph. Decide the channel and the voice before you open a model. Crop or pick the file that will sit next to the copy: the hero packshot, the on-body shot, or the in-use still. Write IMAGE_CONTEXT and CATALOG_SPECS in a notes doc so the prompt stays stable across GPT, Claude, and Gemini. Attach one still per turn, generate title plus bullets plus blurb, edit invented props and fake specs, then paste into the listing fields. Spot-check the title against the live photo on a staging PDP. Save the prompt next to the product template so the next editor does not invent a new voice. Prep and generate-and-paste sit on separate cards so you can hand the same sheet to a contractor.

Batch work follows the same loop with one extra rule: one CHANNEL and one BRAND_VOICE per batch. Mixing Amazon bullets and spare Shopify blurbs in a single thread teaches the model to blend length and tone. Start a new chat when the channel changes. For a 200-SKU drop, lock the skeleton, swap IMAGE_CONTEXT, CATALOG_SPECS, and the file, and keep a spreadsheet column for the edited title you shipped. Models drift if you chat about lighting between SKUs. Soft help if the skeleton is messy: shape the ask once at https://promptmake.net/text, then fill CHANNEL and CATALOG_SPECS yourself.

Prep the file and the catalog facts

1. Open the listing template (Amazon, Shopify, Etsy, or custom PDP). Note title length caps and whether bullets live in a separate field.
2. Pick the hero photo that will sit beside the copy. Crop to the published frame. Watermarks, browser chrome, and collage gutters become fake subjects.
3. Write IMAGE_CONTEXT from sources you trust: SKU title, colorway, view. Do not pull claims from the filename.
4. Paste CATALOG_SPECS: size, capacity, materials, care, compatibility, certifications you hold. Mark unknown where a fact is missing.
5. Set CHANNEL, BRAND_VOICE, and TITLE_MAX. Set PAGE_LANGUAGE if the store is not English.
6. If you hold more than one angle, pick one hero for the first generate. Extra angles go in a follow-up turn with the same skeleton if a bullet needs a back view or an interior shot.

Prep is the part teams skip. A model cannot know your SKU name from a silent packshot. A model cannot know a 12-hour battery unless you say so. Five minutes here save a rewrite after QA flags a fake leather claim.

Generate, edit, and paste into the listing

1. Attach the file to GPT-5.6 Sol, Claude Fable 5, or Gemini 3.5 Flash. Paste the filled skeleton in the same message.
2. Read TITLE against the photo and the catalog. Delete invented materials, wrong color names, and extra clauses that burn the character cap.
3. Read each BULLET. Keep a visible or spec-backed feature. Cut lifestyle poetry that the photo does not support.
4. Read BLURB and META. Confirm the query sits in the title and once in the blurb. Drop a second stuffed repeat.
5. Paste TITLE into the listing title field, BULLETS into the bullet or rich-text field, BLURB into the description, META into the SEO description if your CMS has one.
6. View the staging PDP with the hero photo. If the copy names a lid the crop hides, edit the line before you ship.
7. Store the prompt, the model name, and the date with the product template. Reuse it for the next SKU in the same CHANNEL.

Guests on PromptMake /text get about three generations per day; a free account raises that to about five. Spend those runs on the skeleton. Skip synonym retries of a weak ask. The dedicated product description use case can analyze a still in product if you want photo-in, copy-out. This workflow stays prompt-first so you can run it in any vision chat you pay for.

Product description prompts for titles, bullets, and blurbs

One skeleton covers many SKUs if you swap CHANNEL and the success test. Marketplace titles need searchable nouns and a color or size. Bullets need a feature plus a use. PDP blurbs need two to four sentences that a crawler can index without a slogan. Meta lines need one honest query and a fact. Write a kit per family and keep IMAGE_CONTEXT and CATALOG_SPECS as the moving parts. These kits are paste-ready TASK add-ons. Keep ROLE, FORMAT, and RULES from the skeleton so length and tone do not drift. If a kit and the photo disagree, trust the photo and your catalog, then fix IMAGE_CONTEXT.

Do not mix kits in one turn. An Amazon title kit that leaks into a spare Shopify blurb writes a stuffed 200-character line into a short PDP. A lifestyle kit that leaks into a spec-led industrial SKU writes mood instead of steel grade. New chat, same ROLE, new CHANNEL. Save three snippets in your notes app: marketplace title-and-bullets, web PDP blurb, short meta. That set covers most catalog sites.

Marketplace titles and feature bullets

Title TASK add-on: "Lead with the product type and the SKU name from IMAGE_CONTEXT. Add color or material you can see or that CATALOG_SPECS states. Add one differentiator (capacity, size, pack count). Stay under TITLE_MAX. Skip a slogan."

Good Amazon-style title: Ceramic Travel Mug, Matte Black, 350 ml Double-Wall, Brushed Metal Lid.

Weak title: buy cheap mugs online fast shipping best ceramic cup 2026.

Bullet TASK add-on: "Write five bullets. Each bullet: one feature (visible or from CATALOG_SPECS) then one buyer use. No emoji. No all-caps. No fake reviews."

Good bullet: Double-wall ceramic body; keeps drinks hot on a desk commute without a wet ring on the lid you see in the photo.

Weak bullet: Premium luxury vibe for your best life and morning ritual.

On-body apparel title: Merino Crewneck Sweater, Navy, Crew Fit, Women XS to XL.

Pair or set title: White Leather Sneakers, Extra Laces Included, Unisex Sizes 6 to 12.

Etsy title add-on: "Name the handmade or material fact from CATALOG_SPECS if you verified it. Skip Amazon keyword stuffing. Stay near 140 characters."

PDP blurbs and short SEO meta

Blurb TASK add-on: "Write two to four sentences. Sentence one names the product and the primary use. Sentence two names materials or construction from the photo and CATALOG_SPECS. Later sentences cover care or compatibility. Include the target query once if it fits. Skip buy-now."

Good blurb: This matte black ceramic travel mug holds 350 ml and uses a double-wall body with a brushed metal lid. Take it on a desk day or a short commute. The ceramic exterior matches the packshot; the lid screws on. Dishwasher-safe per the care line in CATALOG_SPECS.

Weak blurb: The perfect mug for anyone who loves coffee and great design.

META TASK add-on: "One line, 140 to 160 characters. Include the product type and one spec. No call to action."

Good meta: Matte black ceramic travel mug, 350 ml double-wall, brushed metal lid. Desk and commute use. Dishwasher-safe.

Weak meta: Shop the best travel mugs on sale now with free shipping today.

Shopify and custom PDPs use a shorter title than Amazon. Set TITLE_MAX to 70 or 80 for nav and collection cards. Keep the long keyword string for Amazon TITLE_MAX. Do not paste an Amazon title into a collection grid and hope the layout holds.

Common mistakes that fail shoppers and search

The frequent failure is stuffing a query into every field. "buy running shoes cheap" in the title, the bullets, and the blurb fails the shopper who reads the PDP and trains search to distrust the listing. Name the shoe you can see, the colorway, and the size range from the catalog. Search can take the nouns that belong to the product. Another frequent failure is asking the model to describe the photo and pasting that caption into the description field. Captions name marble and a mug. Listings name capacity and care.

Other traps show up in audits:

• Inventing leather, stainless, merino, or RFID the photo and CATALOG_SPECS omit
• Using the filename (DSC_4412.jpg or mug-final-v3) as the title
• Writing Midjourney or FLUX camera language into the blurb because a vision chat defaulted to generator dialect
• Claiming a smile, a rank, or a sale the pixels do not show
• Copying one blurb across 40 colorways so every PDP reads the same aside from a swapped adjective

Repair path: classify CHANNEL, fill the skeleton, attach one hero file, paste CATALOG_SPECS, edit invented props, cap TITLE, paste. Change IMAGE_CONTEXT when the colorway changes. Do not fix a thin listing by adding more slogans. Honest specs are the fix.

Scaffold listing prompts in PromptMake /text

PromptMake /text turns a rough listing ask into a labeled prompt you paste into ChatGPT, Claude, or Gemini with the photo attached. You type a messy sentence: "Amazon title, five bullets, short PDP blurb from mug photos, no invented specs, include 350 ml from my sheet." The enhancer returns structure: role, constraints, format. You add CHANNEL, IMAGE_CONTEXT, and CATALOG_SPECS. You attach the still in the vision chat. Soft start: https://promptmake.net/text

Guests get about three text generations per day. A free account raises that to about five. Use a run to lock ROLE / TASK / FORMAT once per CHANNEL. Do not burn the quota on synonym loops. PromptMake /text cannot see your Shopify admin, your Amazon Seller Central, or your spec sheet. You edit the copy and paste it yourself.

Photo-in, copy-out without a chat scaffold lives on the product description use case: https://promptmake.net/use-case/product-description. Pair the two if you want a product path for uploads and a /text path for a reusable prompt you share with a contractor. Keep generation prompts (Midjourney, FLUX, Leonardo) in the image-to-prompt articles. This workflow stops at a title, bullets, and a blurb a shopper can read.

FAQ

What is a product description from image?

A product description from image is listing copy you write from a product photo: a title, feature bullets, and a short SEO blurb for a PDP or marketplace page. You attach the still, add catalog specs the pixels hide, and demand a format a merchandiser can paste. The output serves shoppers and search. Edit invented materials before the listing ships.

How do I write product description prompts from a photo?

Attach the image and lead with ROLE / TASK / FORMAT: listing copy for CHANNEL, title plus five bullets plus blurb, no invented specs. Add IMAGE_CONTEXT for the SKU name and CATALOG_SPECS for capacity, size, care, and compatibility. GPT-5.6 Sol, Claude Fable 5, and Gemini 3.5 Flash all follow that shape as of mid-2026. Treat the first reply as a draft you check against the photo and the spec sheet.

Can ChatGPT or Claude write a product description from a photo?

GPT-5.5 Instant, GPT-5.6 Sol, Claude Fable 5, Claude Opus 5, Claude Sonnet 5, Gemini 3.5 Flash, and Gemini 3.1 Pro can describe an attached product photo. They write usable listing copy when you demand title, bullets, and blurb in sentence one and you supply CATALOG_SPECS. A vague ask yields a caption or a generator-style paragraph. Edit the fields against the pixels before they ship.

What should I add if the photo hides specs?

Paste CATALOG_SPECS in the same turn: milliliters, garment size range, battery hours, fabric content, allergens, certifications you hold. Mark unknown if a number is missing so the model does not invent one. The photo supplies color, silhouette, and visible hardware. Your sheet supplies the facts a returns team will defend.

How is this different from product photography prompts?

Product photography prompts tell Midjourney, FLUX, or GPT Image how to draw or restyle a packshot. Product description prompts tell a vision chat how to write listing copy from a photo you hold. Stay on this page for titles, bullets, and blurbs. Use the image-to-prompt posts when the next step is a new still.

How long should listing copy from a photo be?

Aim for a title under your channel cap (Amazon near 200 characters, Shopify nav 70 to 80, Etsy near 140). Write five bullets of one feature plus one use each. Keep the PDP blurb at two to four sentences and the meta line at 140 to 160 characters. Longer poems fail mobile PDPs and invite stuffed queries.

How do I start with PromptMake /text for free?

Open https://promptmake.net/text and describe the listing job in plain words: CHANNEL, title plus bullets plus blurb, no invented specs, include your milliliters or size range. Generate once, fill IMAGE_CONTEXT and CATALOG_SPECS, then paste the prompt into your vision chat with the photo. Guests get about three text runs per day; registered free users get about five. For a photo-in product path, use https://promptmake.net/use-case/product-description after the prompt shape is stable.