A tiny A6 winter painting became a 200 × 150 cm blanket through 180 AI tries

~From Real Life to AI to Real Use~ Traditional Painting - AI - Large-Format Digital Art Products Project Description “Winter Wonderland” is an acrylic painting I originally created on a very small A6-sized surface. The painting was inspired by my love of winter landscapes, mountain hiking, and rivers. It depicts a peaceful river surrounded by evergreen trees, with three mountain peaks in the background and the pink light of dusk visible on the horizon. The idea for the project came later: I wanted to explore whether a small physical painting could be transformed into a high-resolution digital artwork and eventually into a functional physical product. I photographed the original painting with my phone and used the photograph as a visual reference in Leonardo AI. Because the source image was small and relatively low quality, I experimented extensively with various models, image-to-image, content reference, style reference, elements, prompting, settings, and variations. I created approximately 180 iterations before finding a version that preserved the composition and essence of my original painting while giving it a smooth, polished digital appearance.  I then used GIMP to correct imperfections, remove AI artifacts and noise, make small background edits, and refine the image. Finally, I used Upscayl to enlarge the image from approximately 2,500 pixels to approximately 11,000 pixels, making it suitable for large-format printing. The final digital artwork was uploaded to Redbubble, where I transformed the artwork into a 200 × 150 cm throw blanket. The result is a functional art object that I can use in my own home and offer for sale to others. The final image remains strongly connected to my original painting: the composition and artistic concept are approximately 70% based on my original work, while approximately 30% resulted from AI-assisted transformation and enhancement. Throughout the process, I remained actively involved in the creative direction, refining prompts, selecting between generations, editing the image manually, and correcting imperfections. The same high-resolution artwork can also be applied to other large-format products available through Redbubble, including tapestries. For example, the file is technically suitable for a tapestry measuring up to approximately 260 × 220 cm, although I have not ordered that version yet. Achievements · Transformed a tiny A6 (14.5 x 10 cm) physical painting into a 200 × 150 cm functional art object. · Developed a repeatable workflow for transforming traditional artwork into high-resolution digital art. · Successfully used AI to preserve the identity and composition of an existing artwork rather than generating an unrelated image from scratch. · Created a commercially viable digital asset that can be applied to multiple physical products. · Learned how to combine several AI and traditional editing tools into a single creative workflow. · Demonstrated that an artwork that would otherwise exist only as a small physical painting can be expanded into large-scale art and potentially reach other people's homes. · Successfully produced a physical product that exceeded my expectations: the final blanket is smooth, detailed, and free of the visible AI artifacts that affected my earlier attempt at creating a large-format print. Tools Leonardo AI — image-to-image and content generation, visual transformation, experimentation with prompts and settings, generation of multiple variations, and creation of a smoother digital interpretation of the original painting. GIMP — image editing, artifact removal, noise reduction, background corrections, touch-ups, and refinement of the AI-generated image. Upscayl — high-resolution image upscaling from approximately 2,500 pixels to approximately 11,000 pixels. Redbubble — large-format printing and conversion of the finished digital artwork into a physical 200 × 150 cm throw blanket and other potential products. Methodology The workflow consisted of: Physical painting → Photograph → Leonardo AI → GIMP → Upscayl → Redbubble → Physical product 1. Create the original artwork: I started with a small acrylic painting (A6 page, ~14.5 x 10 cm) inspired by winter mountain landscapes. 2. Digitize the painting: I photographed the physical painting using my phone and used the photograph as the visual reference for the AI workflow. 3. AI transformation: I imported the image into Leonardo AI and experimented with image-to-image, content reference, style reference, generation, prompts, and settings. My goal was not to replace the original artwork, but to create a smooth, high-quality digital interpretation while preserving its composition and artistic identity. 4. Iterative refinement: I generated approximately 180 variations and continuously adjusted my prompts and settings based on the results. One of the main challenges was achieving a smooth digital appearance without introducing excessive noise or unwanted AI artifacts. 5. Manual refinement: I brought the selected version into GIMP and manually corrected imperfections, removed artifacts and noise, adjusted parts of the background, and made small visual corrections. 6. Upscaling: I used Upscayl to increase the image from approximately 2,500 pixels to approximately 11,000 pixels, making it suitable for large-format printing. 7. Physical application: I uploaded the finished image to Redbubble and selected a large throw blanket as the first final application. 8. Real-world validation: I ordered the blanket and evaluated the physical result. The final product successfully reproduced the smoothness and detail of the digital artwork, with no noticeable artifacts. The result was beyond what I initially expected. Learnings This project taught me that effective AI image creation is not simply about writing a prompt and accepting the first result. Creative direction, iteration, and refinement remain essential. I learned how to: · Develop and optimize prompts for a specific desired visual outcome. · Adapt prompts to work with an existing image rather than starting from a blank canvas. · Understand how different settings and iterations affect the generated result. · Combine generative AI with traditional image-editing software. · Identify and manually correct AI-generated artifacts. · Upscale relatively small source material into images suitable for very large physical applications. · Use AI as an extension of traditional artistic skills rather than as a replacement for them. The project also changed how I think about my physical artwork. A small painting no longer necessarily has to remain a small painting. Generative AI can act as a bridge between traditional art and digital creation, allowing an artist to expand, reinterpret, upscale, and adapt existing work into entirely new formats. At the same time, I learned that AI workflows are not completely predictable. Even when using the same model and a carefully developed workflow, results can vary. Models are continuously updated, and a workflow that produces a particular result today may behave differently in the future. There is therefore no completely guaranteed “100% reliable” workflow; successful AI creation still requires experimentation, judgment, and human creative decisions. Perhaps the most important learning was discovering that I could take something very small and personal and turn it into something much larger and tangible. The original painting was only A6-sized 14.5 x 10 cm, but it ultimately became a 200 × 150 cm functional artwork that can bring the atmosphere of that winter landscape into someone's home. Relevant URLs Redbubble product: https://www.redbubble.com/i/throw-blanket/Winter-stillness-by-iuande-art/181808576/vzy2 Attachments: · Original photo of the painting (A6 size) · AI + GIMP output picture · Final 200 × 150 cm blanket picture · Video showing the intermediary steps and finished product