更新记录

Setting your store's base location just got easier — type your address or business name in the new search box and the pin jumps straight there and saves. No more scrolling and zooming the map to find your street. The search box is available on both the Delivery Map setup page and the Base Location page in your company settings. Clicking the map and dragging the pin still work as before, and we've also fixed an issue where dragging the pin on the Delivery Map page didn't save the new position. Since every delivery fee is measured from this pin, it's worth a quick check that yours sits exactly on your store.

Your store's location now also appears as a marker on the Delivery Area map, so you can see exactly where your store sits while you draw the area you deliver to. It updates the moment you move your base location.

Delivery Map fees are now calculated from the actual driving distance between your store and the customer's address, instead of a straight line on the map. If the real route detours around a river, highway, or one-way system, your per-kilometre or tiered pricing now reflects it — and your fees line up with what courier services like Lalamove quote for the same trip.

No setup needed, and your pricing settings are unchanged. Since driving distance is usually a little longer than the old straight-line measurement, it's worth a quick glance at your per-km rate or distance tiers to make sure they still feel right.

Distance-based delivery pricing is no longer capped at five ranges. In the Delivery Map settings you can now add as many "up to X km → price" tiers as you like, and remove any you don't need — no more leaving blank rows or squeezing your zones into a fixed set of five.

A couple of related improvements came along with it:

• One pricing method, clearly chosen. You pick either cost-per-kilometer or distance tiers — the storefront and the cart can no longer show one fee and charge another.
• No accidental free delivery. If an address is farther than your last tier, it's charged your top-tier price instead of slipping through to a zero fee.

Setting up your area is cleaner too: the base location and delivery area are now separate, clearly labelled steps, with a Use current location button and an automatic zoom to your saved zone. Your existing ranges carry over exactly as they were — nothing to redo.

Your storefront's cookie banner now does real consent management, not just a notice. When you turn on Cookie Consent, analytics and marketing trackers — Google Analytics, Facebook Pixel, TikTok Pixel, and any tracking scripts you add in your template's custom head markup — stay switched off until the visitor chooses. The banner offers Accept all, Reject all, and Customize (so visitors can allow analytics but not marketing, or vice versa), and a small Cookie settings button stays on the page so anyone can change their mind later — exactly what GDPR expects. You can also add your Privacy policy URL so a "Learn more" link shows in the banner. Turn it on in Settings → Apps → Cookie Consent; to gate your own custom scripts, tick the new "contains tracking scripts" option in your template's appearance settings.

PromoCards just got more flexible. A card no longer needs an image — leave it out for a clean text card, with an optional custom background colour behind the text (pair a dark background with light text for a bold panel). And a new Join cards (no gap) group setting lets cards sit flush as a single block, with rounded corners only on the outer edges. Put a text-only card next to an image card in a 2-column group, switch it on, and you've built a landing-page-style banner — text on one side, picture on the other — right inside your storefront. Open any group or card and the new options are already there; nothing to set up.

Date Selection popups now offer a Next X Days layout — a clean horizontal strip of upcoming days, with today and tomorrow clearly labelled, that customers tap to pick a date. You decide how many days to show, and the strip scrolls if there are more. Every popup calendar — the week, bi-weekly, and month views included — now highlights the chosen day in your storefront's theme color, so the picker always matches your brand.

You can now add a Next X Days field to your items and to your checkout form — a clean horizontal strip of upcoming days the customer taps to choose, instead of a full calendar. Choose how many days to show, set an optional order cut-off hour, and it can automatically skip the days you're closed. Today and tomorrow are clearly labelled, and the selected day uses your storefront's theme color. It's the friendlier choice whenever you take orders or bookings only a short window ahead.

Your checkout date/time picker can now show a clean horizontal strip of tappable date cards instead of a full calendar — the same friendly style as the Next X Days field. Pick Date cards for a date-only chooser, or Date cards & predefined time slot to pair it with your fixed time windows shown as tappable cards too. Today and tomorrow are clearly labelled, the days you're closed are greyed out automatically, and the selected day uses your storefront's theme color. It's the mobile-friendly choice when customers order a short window ahead — and the full calendar is still there for longer booking horizons. Set it under Settings → Time Slot → Picker type.

Two upgrades shipped together for PromoCards this week.

1. Group heading and description. When you create or edit a group, two new optional fields sit just under the internal group name: Heading and Description. Whatever you type renders on the storefront right above that group's cards — bold heading, muted lead paragraph — the same way category titles and descriptions read on the menu. Use them when a group needs framing: a Today's specials row reads better with a "Available until 10pm tonight" lead; a category-bottom row of related services can introduce itself with "You may also like" before the cards. Leave both blank and the group renders exactly as it always has. The internal Group Name field is unchanged — that one is still for your own reference in the group list.

2. Every customer-facing field is now translatable. On multi-language storefronts, the group heading and description plus each card's Title, Subtitle, Description, and Button Text all support per-language values. Switch language with the top-bar selector, fill the fields in that language, save. Visitors automatically see whichever language they're browsing in, and any untranslated field falls back to your default language so cards always render. Images, button URLs, layouts, colours, and schedules stay shared across languages — change them once and they apply everywhere.

No migration steps for you — open any existing group or card, the new fields are right there.

You can now drop a pin on the map for each pickup area, so customers see exactly where to collect their order. At checkout, areas you've pinned show a small map and a Get directions button that opens the customer's phone maps app.

You can now upload item photos straight from your phone. Large images are resized automatically, so uploads no longer fail for being too big — no need to shrink them first.

If your storefront runs in more than one language, every text field on a Carousel slide now translates per language: Title, Subtitle, and Button Text. Switch language using the top-bar selector, fill in the text for that language on a slide, and save. Visitors automatically see whichever language they're browsing in, and any untranslated field falls back to your default language so slides always render.

What stays shared across languages. Images (desktop and mobile), the button URL, text position, colours, overlay, transition effect, schedules, and the slide order — change them once and they apply everywhere. Only customer-facing text is per-language.

No setup needed — open any existing slide and the fields are already wired up.

Length counters now show on the Title, Subtitle, and Button Text fields too, matching the rest of the admin.

If your storefront runs in more than one language, the announcement bar's Message and Link Text (used by Button Link and Text Link actions) now translate per language. Switch language using the top-bar selector, fill the fields for that language, save. Visitors automatically see whichever language they're browsing in, and any untranslated field falls back to your default language so the bar never breaks.

What stays shared across languages. The link URL, action type (Countdown / Coupon / Button / Text Link), schedule, colours, and coupon selection are shared — change them once and they apply everywhere. (The coupon's own name and discount text translate inside the Coupons module.)

Link Text now caps at 60 characters (was 100). 60 fits long-form German/French CTAs without letting merchants ship a wall of text on a button. Existing entries longer than 60 keep displaying fine on the storefront but will need trimming the next time you save them.

No setup needed — open any existing announcement and the fields are wired up.

When you remove a language from your storefront, WhatsMenu now copies any text that only existed in that language across to your remaining default — your category names, item names, descriptions, summaries, and company info stay intact.

Before: if you'd added a second language but only typed translations for some fields, deleting your original language could leave the storefront showing empty names or descriptions on the other side.

Now:

• Removing a language fills any blanks in your new default from the language you're about to drop.
• Translations you've already written in the surviving language are kept exactly as they are — nothing is overwritten.
• Adding a new language also seeds it from your current language more completely, so you start with a full catalog to translate from instead of empty fields.

Nothing to enable. Applies to every storefront from this update onwards.

If you run your storefront in more than one language, your Custom Item Fields — the merchant-defined extra fields per item (size, vehicle make, service type, special instructions) — now translate alongside everything else customers see.

In the Fields Editor. Open any item, click Open Fields Editor, and you'll see a row of language tabs above the field list — one per menu language you've enabled. Click a tab to switch which language slot every input reads from and writes into. Type the translated label, placeholder, and select-option values; switch back to confirm. Your previous-language values are preserved when you switch tabs.

Single-language stores see the editor exactly as before — no language tabs appear.

AI auto-translate covers them too. When you tick AI auto-translate to other menu languages on item save, the AI now fans the custom-field titles, placeholders, and select options out into every enabled menu language alongside the item's name, summary, description, variant group names, and add-on names. Same credit cost as the other translatable fields. Hand-edited translations are preserved unless you also tick Force translate all.

Existing fields keep working. Whatever language you'd typed your custom fields in before now lives in that language's slot — customers in other languages see your original text until you (or AI auto-translate) fill in their language. No customer-facing breakage; the storefront resolves to whichever language is configured.

What's not in scope (yet). Order-level custom fields configured under Settings > Checkout > Custom Fields stay single-language for now — that surface uses a separate editor and ships in its own slice. The per-item custom fields covered here are the ones on the Custom Fields card on each item's edit page.

Every AI feature in the panel — AI Enhance on items, Draft with AI on orders, the three AI Insight advisors (Catalog, Timing, Restock Forecast) — now renders output as the AI generates it. Tokens appear as they're written rather than landing all at once after a 5-25 second wait.

The total wait time is the same as before; what changed is what you see during the wait. Same trick the big consumer chat tools use — output looks like it's being typed for you in real time, so you can start reading and judging the result well before generation finishes.

Three small bonuses that come with the change:

• Bail-out saves credits. Close the AI Enhance modal (or click Generate again) while it's still streaming, and we cancel the request before the full charge lands. Useful when the first sentence is clearly off-brand and you want a fresh take.
• Long-form output feels alive. The Catalog Advisor digest emits 3 paragraphs of analysis — about 1500 tokens. Watching that build paragraph by paragraph beats watching a spinner for 15 seconds.
• Re-renders cleanly. During streaming you see plain text (so half-formed Markdown markers don't flicker). When generation finishes, the description re-renders with proper bold, lists, and links — same visual as before.

What you don't need to do: nothing. The change applies automatically. Same credit cost per generation. Cache hits (re-saving without changing the source) are still free and now arrive instantly as one chunk.

One thing that hasn't changed: the AI auto-translate background pass on item save still runs silently after your save — no UI to stream into. Refresh the item page a few seconds later to see translations.

Two new AI tools sit on every item's edit page next to the image upload field: Generate image with AI (creates a fresh image from your item's name and description) and Enhance photo with AI (cleans up the photo you already uploaded — background removal, lighting fix, sharper focus).

When to use which:

• You have a real photo → Use Enhance. Photoroom-grade cleanup of your own snap. The result still depicts your real product, so no customer-facing AI label is added.
• You don't have a photo yet → Use Generate. Fills the slot with an AI-rendered image based on the item name and description you've typed. The storefront shows a small "AI-generated" badge on the item card so customers know upfront.
• You want to try a different style on an AI image → Use Generate again. Each click opens the preview modal and lets you refine the prompt ("warmer lighting", "outdoor setting") and try again.

How it works:

Click the button → a preview modal opens with your current photo on the left (if any) and an empty slot on the right. Click Generate or Enhance → the AI processes for a few seconds → the result appears in the right slot. Like it? Click Use this image to save. Want a different angle? Click Try again — each attempt is a fresh AI run.

The preview-then-commit flow means an attempt you don't like never lands on your storefront. You see the result first, then decide.

What it costs:

5 credits per attempt. Each click on Generate or Enhance charges your credit balance, whether or not you save the result. The platform pays OpenAI / Photoroom for every generation; the credit covers that cost. The modal shows a running count: "Attempt 3 — 15 credits spent so far."

If you'd rather not spend platform credits, you can plug in your own OpenAI API key and (for enhancement) your own Photoroom API key in Settings → Apps → AI Kit. With either key on file, the calls route through your account, your provider bills you directly, and platform credits aren't deducted. See the BYOK update for details.

The customer-facing badge — and why it matters:

Items whose photo was created by Generate (not photographed) show a small "AI-generated" label in the corner of the item card on your storefront. This badge:

• Protects you from "not as described" customer disputes. Stripe and PayPal both accept this as a chargeback reason. A visible AI label on the storefront establishes that the photo is illustrative — Stripe's dispute reviewers see it and side with you.
• Sets customer expectations up front. Fewer surprises when the dish or product arrives = fewer disputes at the source.

Items enhanced from a real photo show NO badge. Enhancement just cleans up your phone snap — the product depicted is still your real item, so no disclosure is needed.

The badge appears on the catalog grid and on the item detail page. It's small, sits in the top-right corner of the image, and includes a tooltip on hover ("This image was generated by AI from the item description, not photographed.").

Tips for better results:

• Generate: write a clear, descriptive item name and a short description before clicking. The AI builds the image from what you've typed. "Beef burger with cheddar, lettuce, tomato on a brioche bun, side of fries" produces a sharper image than "Burger."
• Enhance: shoot from above on a plain surface for catalog-grid uses; a busy background is harder to clean up perfectly. The AI works best when your phone snap is already in focus and well-lit — Enhance polishes, it doesn't rescue.
• Try again is cheap. Each attempt costs 5 credits; if the first result misses the mark, refine the prompt and try once more. A soft cap of 10 attempts per modal session protects your balance from accidental click-spam.

What's NOT supported yet:

• Bulk Generate / Enhance for the whole catalog. Single-item only for now. Reach out via the Helpbot if this would unblock a catalog rebuild.
• Faces and people. OpenAI's image policy declines portrait-style outputs for safety reasons. Stick to product photography use cases (food, retail, services) — a salon merchant who wants "haircut model" images is better off using real photos with consent.

You can now plug in your own OpenAI API key so AI calls from your company route through your own OpenAI account directly. Your tokens get billed to OpenAI; platform credits aren't deducted for those calls.

Where to set it. Open Settings → Apps → AI Kit (or https://whatsmenu.page/getapp/aikit). A new Bring Your Own Key (BYOK) section sits at the top of the form. Paste your OpenAI key (the one that starts with sk-, created at platform.openai.com/api-keys), click Save Settings.

The key is encrypted at rest and the panel never displays it back to you — the field always renders blank after saving.

To replace your key. Paste the new value and save. It overwrites the previous one.

To remove your key. Tick Remove my stored OpenAI key on next save and click Save Settings. Platform credits resume on the next AI call. (Leaving the field blank without ticking the remove box keeps your current key — no accidental wipe.)

Cached results stay free. When another merchant on the platform has already run the same prompt, your call returns from cache instantly — same as before. Neither your OpenAI account nor platform credits are charged for cache hits. Only fresh generations bill your OpenAI account.

If your key gets rejected by OpenAI (wrong, expired, revoked) the AI feature you triggered surfaces a clear error pointing you back to AI Kit settings to update the key. There's no silent fallback to platform credits — your bill stays predictable.

Covered AI features. Every AI Kit surface that uses chat completions — AI Enhance, photo-to-description, Reply Assist, AI Insight advisors (Catalog, Timing, Restock), AI auto-translate. All of them honour your BYOK key automatically when one is on file.

If you picked Landscape or Portrait under Settings > Apps > Item image, your storefront catalog now shows item cards in that shape — instead of cropping every photo back to a square.

• A clothing-on-hanger shot stays tall.
• A long platter or banner-style image keeps its width.
• Square remains the default and is unchanged.

The shape applies to the whole storefront — every card uses the same aspect so the grid stays tidy. Items uploaded under an earlier setting are cropped (not stretched) into the new shape, so your catalog still looks uniform without re-uploading anything.

No setup needed. If you haven't picked an image size yet, your storefront keeps its current square layout.

A new Draft with AI button sits inside the WhatsApp Customer card on every order. Click it, and the AI writes a short, ready-to-send message right into the same textarea you'd use for a saved template — no copy/paste, no tab switching. You still pick when to send and what to send.

Reply Assist reads the order in front of you and picks the right tone:

• A heads-up that the order is being prepared, or that it's on the way
• An out-of-stock note asking the customer about a substitution
• A warm "thanks for ordering again" for returning customers
• A polite ETA-update for orders running long

You don't pick the scenario — the AI does it from the order data. Each draft costs 1 credit. Read the draft in the textarea, edit if you want, then click Send Whatsapp message as usual.

This is meaningfully different from the "AI" customer messaging you'll see on most other platforms. Most of them are mail-merge in disguise — pick a canned template, fill in {customer_name}, send. Reply Assist generates a new message every time, tuned to your specific order:

• The wording for a returning customer's substitution is different from a brand-new customer's first-order substitution.
• A 50-minute-old "preparing" order gets a more apologetic ETA tone than a 15-minute-old one.
• If you run your panel in Bahasa Malaysia or Chinese, your draft comes back in that language too — same quality of natural, polite copy.

The AI is strictly forbidden from inventing things it doesn't know: it won't claim an item is in stock when stock tracking says it isn't, won't quote prices the system didn't give it, won't promise specific delivery times. If you spot it making any of those up, that's a bug — let us know.

Nothing leaves your account until you send it. Reply Assist is a writing helper, not a chatbot. No autonomous sends, no inbox watching, no bulk broadcasts.

The button shows up automatically inside the WhatsApp Customer card for Owners, Staff, and Managers. Drivers don't see it — dispatching is their concern, customer messaging isn't. The WhatsApp Customer card must be turned on (Settings → Apps → SendStatus) — without it, there's no send button to go with the draft.

If your credits run low, an inline Top Up button appears instead of a draft. If your platform isn't set up for AI yet, your administrator can enable it in settings.

With Reply Assist, all three of this season's AI advisors are now live: Catalog Advisor, Timing Advisor, Restock Forecast, and Reply Assist. They're independent — use whichever one matches the question you have.