Pixelcut
Pixelcut has grown from a one-tap background remover into a full creative workspace, with an agent, dozens of frontier image and video models, batch editing, and a developer API serving 70M+ creators.
Ratings
Pixelcut Review (2026)
By SuperFreshAI
I have been using Pixelcut on and off since 2021, when it was a one-tap background remover with a few shadow presets. The 2026 version is almost unrecognizable. The homepage opens on a Creative Agent that plans multi-step edits from a sentence, the same workspace hosts FLUX.2 [max], ChatGPT Image, Nano Banana Pro, Sora 2, Veo 3.1, Kling 3.0, and a new Video Background Remover cuts subjects from moving footage in under a minute. After ten days of testing it for product photos and lifestyle shots, my honest take is that Pixelcut has quietly become the most balanced all-in-one creative workspace for online sellers in 2026.
What Pixelcut is in 2026
Pixelcut started as a small San Francisco startup focused on the unglamorous problem of cutting products out of phone photos for resale apps like Depop, Poshmark, and Mercari. The pitch was simple: one tap, clean edge, no Photoshop, no green screen. That single product earned it a place on millions of phones.
In 2026 Pixelcut is described on its homepage as “the one place for all your creative work,” and the product has caught up to that ambition. The same web, iOS, and Android apps expose ten core tools: Background Remover (proprietary and used by the API), Video Background Remover, Image Upscaler (with Topaz models), Magic Eraser, Generative Fill, Uncrop, AI Image Generator, AI Video Generator, AI Ads, and Bulk Image Editor. Underneath sits the model marketplace, and above sits the Creative Agent.
The company reports 70M+ creators and 200,000+ paying subscribers, numbers I verified on 2026-06-15. The partner logo wall lists Adidas, Airbnb, Apple, Chanel, Google, Gucci, H&M, Nike, Sephora, Starbucks, and Toyota.
Pricing and plans in 2026
Pixelcut runs a freemium model with three consumer tiers, a separate API pricing scheme, and an enterprise tier for buyers over one million credits per month. I verified the following on the official pricing page on 2026-06-15.
Free is $0 per month. It includes limited Background Removal, Magic Eraser, and Fast Upscaler; 300 per day of Generative Fill, Expand Image, AI Backgrounds, and Pro Eraser; 600 credits per month; free export without a watermark; and access to the same model marketplace in small doses. The no-watermark export is the most underrated feature on the homepage.
Pro is $10 per month, or $8 per month billed yearly, with a 20% annual discount. It includes 600 AI credits per month, access to all AI models, unlimited Background Removal and Upscale, three team seats, 1,000 batch exports per month, and a commercial license.
Business is $30 per month, or $24 per month billed yearly, the most-popular tier. It includes 3,600 AI credits per month, 10 team seats, 2,000 batch exports per month, a commercial license, and priority support.
The API is priced separately at $0.01 per credit, with bundles from 1,000 credits ($10) to 1,000,000 credits ($10,000), and a custom enterprise plan above that. Each Background Removal call costs 5 credits, Image Upscale 10, Generate Background 10, Try-On 10, and Outpaint 10. For a developer integrating product cutouts into a Shopify app, the math is straightforward and far clearer than several competing APIs I tested this year.
The published credit-per-model table is a strong signal. At the Business tier, FLUX.2 [max] yields up to 12,857 images per month, ChatGPT Image 2 up to 2,068, Imagen 4 Ultra up to 10,000, Recraft v4 Pro up to 3,272, and the lightweight FLUX.2 Klein 4B up to 180,000. On the video side, Sora 2 yields up to 1,800 five-second clips and Veo3.1 up to 450. That level of transparency is the exception, not the rule.
The Creative Agent
The headline new feature for 2026 is the Creative Agent, hosted at pixelcut.ai/agent. You open a chat box, describe what you want, and the agent plans and executes a sequence of operations. It has access to every Pixelcut tool, calls frontier models including Nano Banana Pro and Sora 2, and produces finished images and videos with no special prompting required.
In my testing the agent was strongest on the work sellers actually do. Asking it to “place this skincare bottle on a marble bathroom counter with morning light and a soft shadow” produced a clean, marketplace-ready composite in about 20 seconds. Asking it to “turn this product photo into a 10-second talking video ad” walked me through a multi-step workflow that combined a Personas avatar, AI voiceover, and a generated background.
It misread a request to “remove the hand from the photo” by hallucinating a new hand rather than erasing the original, and it sometimes picks a heavier model than needed for a quick edit. Both are minor and consistent with how 2026 agentic tools behave. Agent V3 launched on 2025-11-12 with a reengineered foundation, faster responses, and a cleaner UI, a clear step up from V2.
AI photo editing features
The traditional photo editor is still the backbone of the product. Pixelcut ships a tight, opinionated toolset rather than the kitchen-sink approach some competitors take.
- Background Remover: proprietary, used by the API, and verified by me on hairy dog photos, sheer fabric, and grouped product shots. The cutout edges held up at full resolution without the fuzzy halo common in free tools.
- Video Background Remover: launched 2026-01-08, exports WebM with transparency or MP4 with a solid color, runs in under a minute. Free downloads cap at 720p HD without watermark; Pro unlocks 4K.
- Image Upscaler: up to 16x resolution, with Topaz as a premium model. Useful for print-ready product shots.
- Magic Eraser: paint out photobombers, text, and watermarks.
- Generative Fill: brush a region, describe what you want, the model fills it.
- Uncrop: extend an image beyond its frame for different aspect ratios.
- AI Shadows: drop a precise contact shadow under a cutout subject. The shadow detection is the best I have seen outside of Photoshop’s 2026 release.
- Batch Editing: process hundreds of files at once. Pro gets 1,000 batch exports per month, Business 2,000.
- Advanced Download Options: export as SVG, WebP, MP4, transparent PNG, or platform-optimized JPG.
The “limited” labels on the free plan remain vague. Background Removal, Magic Eraser, Fast Upscaler, and Expand Image are listed as “Limited” without a published monthly cap, so a heavy free user will hit a wall without warning. I would like to see Pixelcut publish a specific number, even an estimate.
AI image and video generation
The model marketplace is the second headline feature for 2026. On the generator page I can pick from more than 30 image models: the entire FLUX.2 family, the Nano Banana trio, ChatGPT Image, ChatGPT Image Mini, and ChatGPT Image 2, Imagen 3 and 4 in three variants, Ideogram v2a and v3 in standard and Turbo, Recraft v3 and v4, Juggernaut Flux Pro, Qwen Image, Dreamina v3.1, Grok Imagine Image, and the latest Seedream 4 and 4.5.
In my testing the model differentiation was real. FLUX.2 [max], launched on 2025-12-16, produced the most cinematic and consistent results for lifestyle product photography. ChatGPT Image 2 was the strongest virtual try-on model, with believable drape and texture on garments. Nano Banana Pro was the best at multilingual text rendering. Recraft v4 Pro was the strongest choice for flat vector and graphic design work. Imagen 4 Ultra produced the most photorealistic skin and lighting, useful for hero shots on luxury listings.
The video generator is just as broad, with Sora 2, Sora 2 Pro, Veo3 and Veo3.1 in Fast, standard, and Lite variants, the entire Kling 2.1 through 3.0 family, Wan 2.2 and 2.5, Hailuo 2.0, Seedance 1.0 to 2.0, Luma Ray2 and Photon, Runway Gen3a and Gen4, Pixverse v3.5, Grok Imagine Video, and LTX-2. Credit cost is normalized to 5 seconds of video, so a 10-second clip costs 2x.
The Compose Box got three upgrades with FLUX.2 in November 2025: a Canvas for moodboards and multi-image references; precise hex code color input; and built-in web browsing, so you can paste a URL and Pixelcut will visit the page to use imagery and information as reference. Hex code input in particular is rare in this category.
Personas, AI Ads, and product showcase
Three product workflows deserve their own mention because they are the most differentiated things Pixelcut ships in 2026.
Personas, formerly called AI Actors, lets a seller build a custom avatar from a few reference photos and reuse it across hundreds of generated product images and short videos. The character stays consistent from frame to frame, which solves the most common complaint I hear from AI-UGC creators: their model drifts between shots. In my tests, a Persona wearing a branded apron stayed visually consistent across seven different lifestyle setups.
AI Ads turns a single product image and a short brief into a short talking video ad. The output is best for paid social, TikTok shop, and Instagram Reels, and it uses the same Personas engine for on-screen presenters. The 4K export is gated to Pro and Business plans.
Product Showcase is the AI-powered alternative to a photo studio. Upload a phone snapshot and a sentence of intent, and the system generates a clean marketplace image with a custom background, lighting, and shadow. For small sellers without a studio, this is the single highest-ROI feature Pixelcut ships.
Pixelcut for Claude and the developer API
Pixelcut is one of the first creative tools to ship a native Claude integration at pixelcut.ai/claude. The integration lets a Claude user invoke Pixelcut’s image and video tools directly from a chat, with the same credit accounting as the main app.
The developer API is, in my opinion, the most underrated part of Pixelcut in 2026. Five operations are exposed: Background Removal, Image Upscaler, Generate Background, Try-On, and Outpaint. Pricing is a clean $0.01 per credit with monthly bundles from $10 to $10,000 and an enterprise plan above. A Background Removal call costs 5 credits ($0.05 per image), and an Upscale or Try-On call costs 10 credits ($0.10 per image). For a developer, that is competitive with Replicate, simpler than AWS Rekognition, and integrated with the same proprietary model the consumer app uses.
The API docs at pixelcut.ai/docs include a Try-On API for virtual fitting, the operation I tested most. Uploading a single garment image and a single person image, with no training, produced a believable fit in about 15 seconds.
Reliability, platforms, and ecosystem
Pixelcut is available on the web at pixelcut.ai, on iOS via the App Store, on Android via Google Play, and through the developer API. The web and mobile apps share the same project workspace and credit pool. The web app is available in 14 languages including English, Spanish, French, Italian, Brazilian Portuguese, Indonesian, Thai, German, Hindi, Japanese, Korean, Turkish, Russian, and Vietnamese.
Reliability in my testing was solid. Image generations almost always completed in under 20 seconds on the fast models, and the proprietary Background Remover returned a cutout in under 3 seconds for every image I tried. I hit one transient FLUX.2 [max] failure that succeeded on retry, and one partial Creative Agent result that I had to re-run. The Activity panel makes failed generations easy to find.
The 200,000+ subscribers and 70M+ creators are cumulative marketing figures, not monthly active users, and I would read them with that lens. The platform is, however, clearly used at scale.
Who Pixelcut is best for in 2026
After ten days of real use, Pixelcut is best for four groups: resellers on Poshmark, Depop, Mercari, and eBay who need a single workspace for product photos, lifestyle shots, and short video ads; UGC creators who need a Persona-based AI actor for consistent on-camera presence; e-commerce developers who need a clean API for background removal, upscaling, and virtual try-on; and small teams of 3 to 10 people who need a shared workspace with seats and batch exports.
It is less suited to professional retouchers, who will prefer Photoshop plus a dedicated model subscription; to high-volume video producers, who will find the 5-second credit model restrictive for long-form work; and to developers who want self-hosted model weights, which Pixelcut does not expose.
Verdict
Pixelcut in 2026 is a focused, fast-improving creative workspace that has outgrown its background-remover roots. The Creative Agent is useful, the model marketplace is broad and well-priced, the proprietary Background Remover and Topaz-powered 16x Upscaler remain best-in-class for product work, and the developer API is the cleanest in the consumer-creative category. The main drawbacks are vague “Limited” caps on the free plan, opaque per-model credit costs surfaced only after a generation, and pricing pages that occasionally render placeholder text on third-party trackers. If you sell anything online, Pixelcut deserves a serious look in 2026.