Back to VeyraPix

What’s new

The improvements you can actually see in VeyraPix, newest first. Changes under the hood are left out.

    • Improved

      Images with sexual content are now refused right at upload, with a clear message. Before, such an image was accepted and the AI only refused it after a few minutes of waiting. Lingerie and swimwear product photos remain welcome.

    • New

      Kling Image 4, for new images and for editing. Renders straight to 2K or 4K, with photographic realism and sharp, legible text in the image, such as labels and packaging. 240 credits per image at 2K, 320 at 4K. Editing takes up to 10 images at once, for 240 credits (4K: 360).

    • Fixed

      Through the VeyraPix connector your AI assistant can now use older images with a very long name as a source again; they were refused. The Agent’s automatic steps show up in the assistant’s list under a plain name ("Cropped") instead of an internal one, and when a source image really can no longer be read, the connector says so clearly and what to do next.

    • New

      Through the VeyraPix connector your AI assistant (Claude, ChatGPT, Cursor) can now do a lot more: turn a product photo into a 3D model (GLB, at the same price as in the app), split an image into separate layers that each land in your gallery, and check for free whether your photos meet the image rules of Amazon, bol.com, Zalando, eBay, Etsy or Google Shopping, with what to do about each finding. The assistant can also put your subject in a new scene, extend the frame of an image (free on a plain background, or with AI for a photo), combine separate layers back into one image and, only after you confirm, delete images from your gallery.

    • Improved

      On your phone the four tabs of Credits & billing now fit the screen in two rows. In the gallery the quick buttons that appear when you hover an image now sit along the bottom of the tile. They were in the middle, so a click meant to open the image often started a tool. In the light theme the labels on images and the date in the history strip are readable again. Every tab now has its own name ("Gallery | VeyraPix", "Retouch | VeyraPix"), so you can tell tools apart in your tabs and history. With the keyboard the first Tab press jumps past the menu to the content, and the focus ring is twice as thick. On your phone language and theme live in the account menu, which makes the buttons in the top bar larger. In the gallery the Agent’s automatic steps are named after what they did ("Cropped") instead of an internal name, and Product Studio shows a shoot’s status in plain words.

    • Improved

      The history strip below the image shows the day ("Nov 13") for older images instead of only a time, so the order makes sense when you look back several days. In the Agent editor each choice under the text box now says what it sets: Style, Model, Ratio and Quality. The text box in New image is taller, and a project without a cover image gets a folder icon on the start screen instead of an empty square.

    • Improved

      On your phone a tool now scrolls as one page: first all the settings at full length, then the image below. The panel used to sit in its own scroll box of less than half the screen. In the gallery your images start higher up, because the tagline, the counter and the compact button stay away on a phone and sorting, filtering and grouping sit behind one button next to the search field. That button turns purple while a filter is on. The smallest text is one step larger there, and the plus and minus buttons and the gallery buttons are easier to hit.

    • Fixed

      On your phone the start screen and Product Studio no longer scroll sideways: the project cards and the list of shoots fit the screen again. The notice that VeyraPix works best on a computer now remembers that you chose "Continue anyway" and no longer returns with every tool; the button itself is bigger too. And the menu bar fades out on the right to show you can swipe on to 3D, Finish and Gallery.

    • Improved

      In the gallery the labels in the top-left corner ("Edited", "Cropped", "Generated") are now readable on white product photos too. Image to 3D says above its button why it is still grey: pick a photo first. And on the start screen the link next to "Pick up where you left off" is now "All images" and opens the gallery, instead of "All projects" opening your last tool.

    • Improved

      Improve quality is calmer: you first pick what you want done (unblur, restore an old photo, make it bigger, remove a watermark…), and the twenty-odd models sit folded under "All models". If you had picked another model before, that list stays open. And an old or mistyped address such as /editor/remove-bg now opens the right tool, or says the page does not exist instead of silently opening your last tool. And /editor really opens the tool you left off in again; it sometimes showed the Agent under another tool’s address.

    • Improved

      A tool’s main button now stays in view at the bottom, even while you scroll through a long list of options: in Background, Improve quality, Crop & size and Layer extractor, as Light & colour already did. The price sits in the same place on the right of that button in every tool, as "free", "N cr" or, in Layer extractor, "up to N cr". In Several at once the small grey text is larger and easier to read.

    • Improved

      Easier on a tablet and with the keyboard. The confirmation dialog (for deleting or restoring) closes with Escape, puts the focus on Cancel and has buttons in your own language. In Crop & size the page no longer scrolls along when you drag the frame with a finger, and the corner handles are easier to hit. Buttons that only appeared under the mouse (quick actions in the gallery, download in the strip of earlier images, render in Draft grid, remove on references and looks) now simply show on a touchscreen.

    • Improved

      The same names everywhere, in your own language. Retouch, Improve quality, Export, Crop & size and Image critique now carry their menu names in their panel, in the gallery and in the Agent, and the mask tools now speak English in the English view. Under each Improve quality job you now read what it does instead of a model name. Errors in Product Studio, Shopify, API keys, Draft grid and Cinematic Shots are plain sentences instead of "Save failed (500)". Downloads from the gallery and the canvas are named veyrapix-… and keep their real file type.

    • Fixed

      A run that got stuck, for example because you closed the tab during a render, no longer holds you up: after fifteen minutes it stops counting towards how many runs you may have going at once. On a plan with one run at a time this could keep you from starting anything for hours, with a message to wait while nothing was running.

    • Improved

      Ask the Agent for something it does not do in the chat and it says so plainly, with a button to the tool that does: a 3D model goes to Image to 3D, a series of more than four shots to Cinematic Shots and a full product shoot to Product Studio. Single product photos and up to four images at a time it still makes itself.

    • Fixed

      The Agent no longer starts a paid edit before you have seen the price. A concept card, "Fix: …" or a follow-up suggestion first shows the card with model and price, unless Auto-run is on. You pay what the card says: in Easy the model on the card is the one that runs, and if you change your settings afterwards you see the new price first. Words like "premium", "high-key" or "high contrast" no longer push the quality to 4K; only "high quality" or "4K" in your own text does.

    • Fixed

      An upscale to an exact size (say 6000×4000) now costs what the card says, and Stop says honestly that an upscale already sent keeps running. If a request fails before it reaches the service, your credits come back right away. "Make them all black and white" no longer starts a batch of old suggestions, a new chat stops a run that is still going, and a step the Agent cannot run is no longer quietly replaced by a paid edit: it asks you again.

    • Improved

      The Agent is friendlier to use. Free steps (crop, light & colour, removing the background on your device, a plain background, export) now run straight away, without a confirmation first. If you are short on credits, it says how many you need and offers a button to top up. Messages are in your own language and in plain words, steps carry the menu names ("Light & colour", "Improve quality"), and the start button says what will happen. Downloads are now named veyrapix-… and keep their real file type, so a cut-out stays .png.

    • Improved

      The model picker in Create is tidied up too: Grok Imagine (old version) and Qwen2 are gone, because their successors Grok Imagine 2.0 and Qwen Image 2.1 are cheaper (80 instead of 120 credits). If you had picked one, you get its successor automatically; older projects still open.

    • New

      In the Agent, tell it what each reference is for. Click the label at the top left of a reference: Ref (general), Style (follow colour, light and mood) or Person (keep this face recognisable). A style or person reference is read first and then applied to your image on purpose.

    • Improved

      The Agent is easier to use. Type an instruction and then draw a mask, and you get the review card with the price straight away instead of a proposal first. Buttons, labels and messages are now in your language everywhere, without technical error texts. On a phone the buttons on a result stay visible and the bars wrap neatly on a narrow screen, and you can pick references with the keyboard too. Clearing memory now asks for confirmation first.

    • Fixed

      Smaller Agent fixes: a workflow always edits its own intermediate result, even if you click another image meanwhile. Retrying no longer charges again for steps that already finished. If a masked edit cannot be placed back, your image stays as it was. Your typed text stays when you press a button, the working brief keeps your edits, and switching quickly to another chat keeps your last change.

    • Fixed

      "Extend the frame" in Upscale now really adds room. Choose the direction (all sides, left and right, one side) and how much room or which format, such as 16:9. Your original pixels stay exactly where they are. A plain or transparent background is extended for free; for a photo a model paints only the new border, for 120 credits. The old model often returned the same picture.

    • New

      The Agent now changes the camera angle of your photo itself. Ask for the back, the side, a view from above or below, or a close-up or wider shot, and it re-renders the subject from that viewpoint, just like Camera Control. You see the price of both renders on the button before you start.

    • New

      The Agent now also runs the specialised Upscale jobs: remove a watermark, restore an old photo, unblur, remove noise and fix the lighting. Say what is wrong with your photo; you no longer need to know which model does it.

    • New

      The Agent now really extends your image. Ask for more room all around, on one side, or to a format such as 16:9, and your original pixels stay exactly as they were. If your product stands on a plain or transparent background, that background is extended for free. For a photo, a model paints only the new border, for 120 credits.

    • Fixed

      Some images in your gallery, mostly Agent results with a long name, could not be picked as a source: clicking them did nothing. They open again now, including the ones you already had.

    • Fixed

      Stop in the Agent now really stops. Press Stop during a set of variants or a workflow and the remaining images and steps are no longer started or charged. Stop before anything is sent and it costs you nothing.

    • Fixed

      Cropping to a ratio such as 1:1 keeps the whole product in frame. If the product does not fit, the Agent extends the plain or transparent background instead of cutting off the toe and heel. Ordinary photos are still cropped from the centre.

    • Improved

      The Agent's approve button now always shows the price in credits, also for a single image. And the workflow card shows what each step does, such as "white · soft shadow", "1:1" or "WEBP", so you know what you are approving.

    • Fixed

      A new Agent chat no longer carries layers over from an earlier chat, the Agent no longer adds an export you did not ask for, and the Point out button opens again.

    • New

      New 3D menu, with Image to 3D: turn one product photo into a 3D model with textures and PBR materials. Choose Hunyuan 3D 3.1 Rapid (990 credits, the cheapest), Tripo H3.1 (1,300 credits, best value), Meshy 7.1 (4,330 credits) or Hi3D v3.0 (9,130 credits, top quality). Rotate the model in your browser, download it as a GLB (usually 3 to 10 MB, light enough for a product page) for Blender or your web shop, and view it on your phone in AR. You see the price in credits up front, and if a model fails, your credits are refunded.

    • Improved

      Seedream 5.5 now also does 1.5K, in Create and in Retouch, for the same price as 1K: 80 credits for a new image, 160 for an edit. You get more than twice the pixels, for example 1536×1536 instead of 1024×1024. When you edit part of your photo, Seedream 5.5 picks 1.5K by itself when that is enough.

    • Fixed

      The Layers panel header is tidied up: "Layers", undo and redo no longer overlap, and the Generate, Text and Layer buttons sit on their own row so nothing gets cut off.

    • Improved

      The model picker in Retouch is tidied up: 28 models instead of 34. Older versions with a newer one at the same price or less are gone: Grok Imagine i2i (now Grok Imagine 2.0, 80 instead of 160 credits), Nano Banana Edit (now Nano Banana 2.1), Qwen2 Image Edit (now Qwen Image 2.1, 80 instead of 120) and Seedream 4.5 (now Seedream 5.5, 80 instead of 120 in Create). If you had picked one, you get its successor automatically; older projects still open. Ideogram Reframe now sits only under Extend and Ideogram Character only under Create, since neither edits your photo. "Recommended" now shows the top four picks instead of a long list.

    • Fixed

      Seedream 5.5, FLUX-2 Pro and Flex (new images), Qwen Image 2.1 and Qwen Image 3 Pro no longer offer 4K. These models top out at 2K: picking 4K gave you a smaller image or an error, and on Seedream 5.5 a 1K image at the 2K price. They now show 1K and 2K only. For 4K, pick for example MiniMax H3, Nano Banana Pro or GPT Image.

    • Improved

      Light & colour now has Exposure, Highlights, Shadows and Tint. Exposure works in stops, like a camera, and lets bright areas roll off softly instead of blowing out to white. Highlights brings back an overexposed wall or sky, Shadows opens up dark areas, and Tint removes a green or magenta cast. Hold Before/after to see the original; Apply now always stays in view. Hue is now called Hue shift and sits with the effects. Your saved looks keep working, and the Agent knows the new sliders too.

    • New

      MiniMax H3, for new images and for editing. Renders 1K, 2K and true 4K in one pass, with no upscaler in between. 2K costs the same as 1K: 40 credits per image, 4K 120. Editing takes up to 16 images at once, for 80 credits (4K: 160).

    • Improved

      Regenerating one layer is much cheaper. It now has its own model choice, Nano Banana 2.1 by default: 120 credits per layer including the cutout. It used to use your Retouch model, often 280 to 400 credits. The button and the Agent now also show what you actually pay; they used to show the model's most expensive price, sometimes over 4000 credits.

    • New

      With Pro, the Layer extractor now lets you draw frames on your photo to point out which parts get their own layer. Only those are cut out; everything else stays on the background. If two parts overlap, give the frames a name (for example "green sneaker") and they come out cleanly apart. Up to 6 frames per run.

    • Fixed

      The model picker is accurate again. The English version still showed Dutch text ("33 modellen gevonden"), the description of plain Qwen Image 3 called it "Pro" and had broken characters, and the pricier Flux-2 model is now called Flux-2 Flex. Its description now says what it is: the detail tier, slower and two to three times the price of the faster Flux-2 Pro. It used to say the opposite.

    • Fixed

      The Layer extractor and Image composer no longer show Retouch's brush bar over your image. It did nothing there.

    • Improved

      Pro in the Layer extractor now runs on Seedream 5.5. Your layers keep the colours of your photo much better, the run is faster, and it costs 30 credits per layer at 1K and 50 at 2K, down from 60 and 100. New is 1.5K: more than twice the pixels of 1K, for the same 30 credits per layer. A run returns at most 7 images (background plus 6 parts) and you only pay for what comes out. The desired-layers field can now stay empty: the model then finds the separate parts itself. The Agent uses the same Pro version.

    • Fixed

      The Layer extractor now remembers whether you picked Standard or Pro and which size, also after a reload. It used to jump back to Standard 1K after every run, so a second run could quietly use a different quality.

    • New

      Background now works in one step: pick what you want to end up with (transparent, a plain colour or a new scene) and one button does the rest, cutting out included when it is still needed. The scene is new: describe a setting or pick an example, and Nano Banana 2.1 puts your subject in it with matching light and shadow, at 2K for 120 credits. Cutting out on your own device stays free.

    • Improved

      Area edits now live inside Retouch, under the "Edit several areas" button, instead of as a separate tool in the menu. The screen now shows which model edits your areas, only shows a price once an area is marked and has an instruction (it used to say "~0 cr"), and the choice between "Each area on its own" and "All in one go" is in plain words. If you had pinned Area edits to your home page, Retouch is there now.

    • Fixed

      Retouch no longer gets stuck when you come from Create. If you had picked a model there that only makes new images, such as Seedream 5.5, Retouch kept it, showed a warning and left the button dead. Retouch now switches to the editing version of the same model by itself.

    • Improved

      Every tool on the home page now shows what it does with a real example: before and after for background, retouch, upscaling and more (point at the card and the after slides across), and a small sketch for the rest. The (i) tells you in two or three sentences what goes in, what comes out and whether it costs credits.

    • New

      Your home page now works for you. Pin your favourite tools with the star, on the home page or in the menu at the top, and they wait for you under My flows, in the order you drag them into (up to 8). Nothing pinned yet? Then that row shows the tools you used most this month. Also new: a Featured row with what is new or cheaper, and at the bottom every tool, with a filter per step and a search box.

    • New

      Seedream 5.5, for new images and for editing. The newest Seedream: sharper and more faithful to your prompt than 5.0, and built to survive edit after edit without the image falling apart. At 2K it costs 120 credits (200 to edit) against 280 for Seedream 5 Pro; at 1K, 80 and 160. No 4K — for that, stay on 5 Pro or GPT Image.

    • Fixed

      Changing the camera angle gives you the new angle again. Since mid-October the final render often came back as a copy of your original photo; it now runs on GPT Image 2.6 Sunburst and follows the angle from the preview. The render costs more as a result: about 520 credits at 2K and 840 at 4K, instead of 120 and 200.

    • Improved

      Extracting layers is much cheaper. The new standard quality, Seedream AI Layers Flash, costs 30 credits per layer at 1K and 2K and runs twice as fast; Pro stays one click away and now costs 60 credits per layer at 1K instead of 160.

    • Improved

      If your connection drops for a moment while a page loads, VeyraPix reconnects by itself instead of showing an error screen.

    • New

      GPT Image 2.6 Sunburst — OpenAI’s newest image model, at every quality from 1K to 4K, both for new images and for edits.

    • Improved

      Every model is now priced at what the provider actually charges. Thirty prices were out of date; many models got cheaper, among them Flux Kontext, Z-Image, Seedream v4 edit and most upscalers.

    • Improved

      Your home page opens on the work you already have, instead of on a row of explanations about tools you already know.

    • New

      Area edits are now saved with your project: image, marks and instruction all come back exactly as they were after a reload.

    • Fixed

      A project made with a model that has since been replaced opens again. This affected work from Nano Banana 2, Seedream v4 and GPT Image 1.5, among others.

    • Retired

      Midjourney, Phota Enhance and Topaz Denoise are gone from the picker. Their providers no longer offer them; existing projects still open.

    • New

      Eleven models added that KIE has put in its catalogue since the previous comparison.

    • Improved

      A pass over the whole interface: the editor, the agent, the account page and error messages all say more clearly what is going on and what you can do about it.

    • New

      Two draft-grid templates: a character lineup and a product turnaround.

    • New

      VeyraPix now works from inside Claude and ChatGPT. The editor’s tools are available there, results appear in an image viewer in the conversation, and you can put files from the chat straight into VeyraPix.

    • Improved

      Connecting takes one click through OAuth, instead of a key you have to copy across yourself.

    • New

      Draft grid: explore several ideas for the same shot in one generation, with a freestyle mode that uses no template. You can reopen a grid from the gallery later and render it then.

    • New

      GPT Image 2.5 Sunburst, Nano Banana 2.1 and Seedream 5 Flash added.

    • New

      Background removal and the free workflows run over every image you have loaded, not just the one on the canvas.

    • Fixed

      An export of several images comes back as one archive instead of as separate downloads.

    • New

      Realism Finish: optics and grain over a generated image, with zoom, a before-and-after comparison, and presets you can save.

    • New

      GPT Image 2.5 added, and set as the model for the standard run.

    • New

      Areas: mark the regions that should change, give each region its own reference image, and retry a region as often as you like without touching the rest.

    • New

      Text is a real layer now instead of a generated picture of text, with opacity, mirroring, duplicating and aligning per layer.

    • New

      Composition guide: draw the layout you have in mind, describe it in one sentence, and keep every attempt.

    • Improved

      A workspace home page and shorter navigation, so there is less hunting for the tool you need.