Magnific
AI creative platform (formerly Freepik) built around the Magnific upscaler — upscales, relights and restyles architectural renders, and turns sketches into photoreal images.
Magnific started in 2023 as a single, very specific tool: a generative AI upscaler, built by Javi López and Emilio Nicolás in Murcia, Spain, that didn't just enlarge an image but invented plausible new detail as it went -- sharper brick joints, grain in timber cladding, crisper glazing reflections. Freepik acquired it in May 2024, and on April 28, 2026 the whole Freepik company rebranded under the Magnific name. So 'Magnific' today means two things: the upscaler architects already knew, and a full AI creative platform around it, with 30+ image, video and audio models, a 250M+ asset stock library and more than 1 million paying subscribers.
For architects, the upscaler is still the reason to be here. Magnific offers two models: Classic, a fast, clean upscale for sharpening compressed or low-resolution images, and the Magnific creative model, which goes up to 10K output and adds two controls that matter -- a Creativity slider (how much new detail the AI is allowed to invent) and a Resemblance slider (how closely the result must stay to your original). Bulk upscaling handles up to 20 images at once on paid plans, and it accepts JPG, PNG and WEBP.
In a practice, that makes Magnific a finishing step rather than a renderer. The typical workflow: produce a quick draft out of Enscape, D5 Render, Twinmotion, a SketchUp viewport or an AI generator, then push it through Magnific to get a print-ready or presentation-grade image without re-rendering the scene at 4x the resolution. For a small studio without a dedicated GPU workstation, that trade -- a cheap fast render plus a cloud upscale -- is often the difference between a competition board that holds up at A1 and one that doesn't.
Around the upscaler, the platform now adds tools that are directly useful in architectural visualization. Relight changes the lighting of an existing image from a text instruction (day to dusk, or 'apply night lighting to this space'); Restyle swaps the interior style of a room while keeping its layout; and Magnific's interior and architecture tools start from a prompt or a hand-drawn sketch and generate an early-stage photoreal render -- the image on this page is Magnific's own sketch-to-render example. There's also a video upscaler for walkthrough clips, and Spaces, a shared node-based canvas where a team can chain these steps into a repeatable workflow.
The platform plugs into the tools designers already use: official plugins cover Photoshop, Illustrator, InDesign, Figma, Affinity, ComfyUI, Microsoft Office and the major video editors (After Effects, Premiere Pro, Final Cut Pro, DaVinci Resolve), and there's an API and an MCP server for Claude and ChatGPT. What's missing, for this audience, is any plugin inside a modeling tool -- there is no Revit, SketchUp, Rhino or Archicad integration, so Magnific always works on exported images, never on the model itself.
Pricing is credit-based and billed as a yearly pool rather than monthly resets. Premium is $14.50/month billed annually ($20 month-to-month) with 240K credits a year; Premium+ is $33.75/month billed annually ($45 month-to-month) with 600K credits and unlimited use of several image models; Pro Starter is $82.50/month billed annually ($110 month-to-month) with 1.5M credits. On Premium+ and Pro, the Classic upscaler is unlimited, but the Magnific upscaling models still consume credits -- Magnific Precision, for example, costs 90 credits per 2K image. Magnific's upscaler page also describes full 10K Magnific upscaling as a Premium+ feature, so architects who mainly want the upscaler should look at Premium+ rather than Premium. There's a 30-day refund guarantee, a commercial AI license on paid plans, and Magnific states that uploaded images are never used to train its models.
Editor’s takeBIM•STACK editor's take: Magnific is the best tool in this catalog for one job -- turning a decent, fast render into a large, detailed presentation image -- and the 2026 platform makes it more useful to architects than it was as a standalone upscaler, thanks to Relight, Restyle and sketch-to-render sitting in the same subscription. The caveat is the same one that has always applied: generative upscaling reinterprets detail instead of preserving it, so window mullions, signage, text and material joints can change subtly between the source and the output. Keep the Creativity slider low and the Resemblance slider high for anything a client will read as design intent, and never use an upscaled image as a record of what was actually modeled.
Getting a first result takes minutes: upload a render, choose Classic or the Magnific model, pick a scale and download. Anyone comfortable with Photoshop will find the interface familiar, and the before/after slider makes it easy to judge a result immediately.
The real learning is in the Creativity and Resemblance settings. Most visualization teams settle, after a handful of test images, on a conservative preset for client-facing design imagery (low creativity, high resemblance) and a looser one for mood and competition images, then reuse those settings rather than experimenting on every render. Spaces lets a team save that chain -- upscale, relight, export -- as a shared workflow, which is where the platform starts paying back the extra setup time.
- Small studios producing presentation and competition boardsTurns quick draft renders into large, detailed images without a high-end GPU or a long re-render.
- Post-processing output from real-time renderers and AI generatorsPairs naturally with Enscape, D5 Render, Twinmotion, Lumion or Midjourney output.
- Early-stage concept imagery from sketchesSketch-to-render, Relight and Restyle help test mood, lighting and interior style before anything is modeled in detail.
- Teams that also need video, stock assets or marketing contentOne subscription covers the upscaler plus video models, a video upscaler and a 250M+ asset library.
- You need renders tied directly to your BIM modelVeras or D5 Render work inside Revit, SketchUp and Rhino and follow the model's actual geometry.
- You need a full production renderer with physically accurate materialsV-Ray, Enscape or Lumion render from the model itself instead of enhancing an image after the fact.
- You only need AI renders from CAD exports, not a whole creative platformMyArchitectAI is focused specifically on turning architectural exports into photoreal renders.
- Pixel-exact accuracy matters more than added detailA traditional, non-generative upscaler won't invent detail that wasn't in the source image.
Is Magnific the same as Freepik?
Yes. Freepik acquired the Magnific upscaler in May 2024 and rebranded the whole company as Magnific on April 28, 2026. It's the same company and account, with a broader AI creative platform.
How much does Magnific cost?
Premium is $14.50/month billed annually ($20 month-to-month), Premium+ is $33.75/month billed annually ($45 month-to-month) and Pro Starter is $82.50/month billed annually ($110 month-to-month). Plans are credit-based, with a yearly credit pool.
Can Magnific upscale architectural renders?
Yes -- that's its best-known use. It upscales up to 10K with the Magnific model (a Premium+ feature, per Magnific's upscaler page) and adds detail to textures, glazing and materials, controlled by Creativity and Resemblance sliders.
Does Magnific change details in my render?
It can. Generative upscaling reinterprets detail rather than preserving it pixel for pixel, so keep Creativity low and Resemblance high when accuracy matters.
Does Magnific work inside Revit or SketchUp?
No. Plugins cover Photoshop, Illustrator, Figma, Affinity, ComfyUI, Microsoft Office and video editors, but not modeling tools, so you export an image first.
Does Magnific use my images to train its AI?
No. Magnific states that uploaded and upscaled images are never used to train or fine-tune its models, and you keep the rights to your results.