Art for real interests.

Research gave us a quality bar.

We researched popular canvas brands to understand the visual qualities that make wall art feel intentional: a strong central idea, a subject someone recognizes or cares about, a clear composition, and enough detail to reward a closer look. That research informed a style direction for original INKRALLY work. We are not publishing the reference brands or their artwork.

The catalog grew around interests and moods rather than one repeated visual trick. Technology, aerospace, cosmic subjects, and quieter natural compositions gave the collection different ways to connect with a buyer. The selected works below show some of that range.

The hard part was getting beyond a promising first image.

AI image generation produced the concepts, but generation was only the start. A piece had to work as a complete composition: subject, type, contrast, spacing, and texture. A striking thumbnail could still fail when viewed larger because of awkward lettering, implausible details, or a weak relationship between the image and its message.

Ethan reviewed the work and gave direction on what to keep, revise, or retire. AI then used that feedback to make targeted revisions. For technical subjects, visual confidence was not enough; details still needed scrutiny. This was an iterative art-direction process, with explicit human selection, rather than accepting every image the model produced.

We built a place to collaborate on the work.

INKRALLY Studio brings the artwork and canvas review into one private workspace. Ethan can move through the collection, inspect the front and edges, rotate a canvas preview, choose how a signature should appear, and leave a note attached to a specific piece. AI can work from those recorded decisions instead of trying to reconstruct feedback from a long conversation.

The distinction matters: selecting a favorite is an art-direction decision; changing a signature is a revision; approving an actual printed sample is a different checkpoint. Studio helps keep those decisions attached to the right artwork. The public preview here demonstrates the interface with selected web artwork and no private feedback.

Upscaling had to preserve the art we approved.

The original generated images were not automatically large-format print masters. We tested image-restoration enlargement and inspected the result for changes to texture, lettering, edges, and small features. More pixels alone did not prove that a canvas would look good.

An early attempt to improve print readiness changed parts of an approved composition. Ethan’s correction was decisive: keep the approved art intact. From that point, enlargement and preparation had to preserve the selected work, with comparisons against the source and separate checks for the visible front and wrapped edges. The production story explains the remaining physical-sample checkpoint without publishing the manufacturing recipe.

A catalog is a set of decisions, not a generation count.

The earlier catalog snapshot contained 155 selected digital artworks. That number describes selected work, not printed stock, proven demand, or pieces with completed physical approval. Later signature revisions and product-gallery updates continued to refine the collection after the initial selection.

A later review batch explored 45 original concepts after Ethan asked for deeper Human, History and Sport collections and a developer-focused direction. His ratings and notes narrowed the batch to 24 selected additions and 21 retirements. AI revised the designs that needed changes, preserved the reviewed sources, and brought the approved selection into the storefront with complete product galleries. The verified September 10 storefront snapshot is 179 works. The new additions still need their separate print-file, wrap and fulfillment checks; digital selection is not physical approval.

The takeaway is that AI made broad exploration possible, while selection, revision discipline, and fidelity checks made the work more coherent. The feedback records guide later work; this process does not mean we trained or fine-tuned a foundation model.

The next exploration asks different questions.

Ethan then asked to explore new styles rather than expand into unrelated products. AI researched popular art approaches and commercial image rights, directed image-generation tools, and constructed original city-map compositions. Thirty concepts across five directions entered private review.

The distinction between inspiration and permission became part of the work. A famous photograph or character can involve rights that generating another version does not automatically resolve. This round explores original subjects and thematic ideas, with real geographic data credited where used. Ethan’s next ratings will decide what deserves refinement; none of these concepts is a new storefront product or a print-approved file.

The first exploration missed the mark. Ethan rejected the round because the work looked visibly generated. AI archived those concepts and changed the test: a smaller group using flat graphic composition, stronger typography, and a real archival photograph in a separately constructed layout. At that stage, the three replacements awaited his judgment. Producing a polished-looking image did not mean the art direction was successful.

Ethan liked the smaller graphic and archival direction and asked for more. AI preserved those three originals and developed 18 additional concepts across the accepted approaches. The retained review audit verifies the new sources and review view; individual selection and production approval remain separate. The storefront catalog was not expanded by this exploration.

After liking the expanded graphic direction, Ethan asked for more vibrant color and street-art exploration. AI developed twelve original private concepts spanning pop-art consumer satire, graffiti lettering and stencil compositions. His enthusiastic response led to twelve more in the same direction, bringing this exploration to 24 concepts across two review rounds. The second-round source audit verified that all 212 earlier Studio artworks stayed intact; browser checks verified the twelve new previews without changing existing feedback. These remain exploratory designs awaiting individual selection, not additional storefront products or print-approved files.

The next category test followed Ethan’s choice of music and his request for a better repeatable city series. AI directed eight original music graphics and constructed four coordinated city designs using outlined typography and credited geographic data. The city specimens share one layout while changing the city, street geometry and color. The retained review audit verifies twelve new private concepts, preserves all 224 earlier Studio sources, and confirms existing feedback was unchanged. This tests the visual direction before individual selection or production approval.

Ethan clarified that the music category needed the established technical editorial treatment, rather than the saturated pop-art approach. He kept four earlier music pieces in Pop Art and approved the city system as a direction to develop. AI then made six new audio-object concepts, added two cities to the accepted layout, and built two alternative city layouts using the same pair of cities for comparison. This correction separates a subject category from a visual style. Retained checks verify twelve new private concepts and preserve all 236 earlier Studio sources and existing feedback; individual selection and print validation remain separate.

Two city systems and selected art become a complete release.

Ethan rated the next rounds, selected the technical music direction, and chose Urban Index and City Field for expansion. Four earlier music graphics stayed in Pop Art. AI extended each selected city system across 25 cities, using credited geographic data while preserving the previously reviewed city compositions. Fifty city pieces joined 51 selected non-city pieces in the release. Expanded city specimens follow the approved direction; they were not assigned invented individual ratings.

The work went beyond making thumbnails. AI preserved the original compositions, prepared faithful enlargements and size-specific canvas layouts, and checked the visible fronts separately from the wrapped sides. Each addition received five product-specific photographic composites showing the actual selected design in styled environments. These are generated product visualizations, not evidence of a physical photoshoot or a printed sample.

The retained release checks verify 101 additions, 505 distinct gallery images, 303 prepared size-specific exports, and working product pages on mobile and tablet. That release brought the storefront to 280 artworks. Original review records remain intact. Publishing these pages and preparing digital files do not authorize fulfillment: provider proofs, physical sample approval and commerce enablement remain separate checkpoints.

Ethan subsequently removed the ten Street Art pieces because the direction did not fit his quality bar. The six technical music pieces moved into Technology, keeping the sound subjects while simplifying the collection structure. The active storefront now contains 270 artworks across eleven collections. The withdrawn originals and review history remain preserved privately; their earlier creation and publication remain part of the experiment’s history.