The sheet is a canvas. The references on it live in folders you already have

A character reference sheet usually starts as a folder problem before it's a drawing problem: turnaround sketches in one place, expression studies in another, a colour script somewhere else, and reference photos for a costume detail buried in a "misc" folder from three projects ago. Building the sheet means finding all of that and putting it in one place a person can actually look at — not generating anything new, just assembling what already exists into something legible.

That distinction matters more than it sounds like it should, because the temptation with a character-specific tool is to expect it to do some of the creative work — suggest a pose, fill in a missing angle, propose a palette. Nothing below does any of that. What it does is take the honest, uneven state a real character folder is usually in — three good turnaround attempts and six that didn't work, a costume reference pulled from a completely different project, a palette someone lifted off a mood board months ago — and get from that state to a sheet without a re-draw or a re-collect first.

Why the folder is usually already right, just unsorted

A character's material tends to accumulate correctly even when it looks chaotic: the turnaround sketches are in the concept folder because that's where concept work goes, the costume photo reference is in a shared research folder because it was pulled for more than one character, and the palette lives wherever palettes for this project already live. None of that is wrong. What's missing isn't organization — it's a way to pull the character-specific subset out of folders that are organized around something else entirely, which is what tagging by name does in the next step, rather than requiring everything to be physically moved into one character-specific folder first.

Step 1: Point one root at wherever the character's material already lives

Whether it's a single project folder or references scattered across a few — concept sketches, photo references for a costume piece, a palette pulled from a mood board — register the folder as it already sits. There's no restructuring required first; the first pass reads whatever's there.

Step 2: Tag by character name first

The single most useful tag on a project with more than one character is the character's own name, applied to every piece of material connected to them — sketches, photo references, palette swatches — regardless of which folder each one happens to live in.

Step 3: Tag by sheet section, not just by character

A second, shared vocabulary — turnaround, expression, palette, silhouette, costume-detail — sits alongside the character tag and stays the same across every character you ever build a sheet for. That consistency is what makes the next sheet faster to build than this one.

Keeping a small, reused vocabulary is the difference between a tag list that answers "what section does this belong to" instantly and one that's grown a new one-off tag per character until none of them mean anything specific anymore.

Step 4: Rate to pick the version that actually goes on the sheet

Most characters accumulate more sketches per pose than end up on the final sheet — three attempts at the same three-quarter view, two competing palettes. A rating pass separates "exists" from "the one that's going on the sheet" without deleting anything.

Rating before tagging further is the faster order for this kind of narrowing — deciding "is this the one" moves quicker than deciding "what do I call this" for every sketch in the folder.

Step 5: Assemble the sheet as a canvas

Placing the rated set onto a canvas is what actually produces the sheet — a file that lays the turnaround, the expressions and the palette out together, rather than a folder someone has to open one image at a time to piece together mentally.

Because Mira-Rook-reference-sheet.coii-canvas is pretty-printed JSON sitting in the project's own folder, it travels with the project rather than living inside an app-managed library. Hand the folder to another department and the sheet opens with nothing to re-export.

Step 6: Attach the notes a tag can't carry

A tag says "turnaround." It doesn't say "the jawline in the third pose is the one that's correct — the others are earlier passes." Where a specific instruction attaches to one image rather than a category, a note on that file carries it more precisely than inventing a tag that will only ever apply once.

Step 7: Reuse the sheet's shape for the next character

Once one sheet exists as a canvas, the next one doesn't start from a blank file. Duplicating Mira-Rook-reference-sheet.coii-canvas, renaming it and swapping its plates for the next character's rated set keeps the section layout — turnaround top-left, expressions along the bottom, palette in the corner — consistent across a whole cast, which matters more than it sounds like it should when several artists are producing sheets that need to sit side by side later.

What this doesn't do

Worth stating plainly, since character-reference tools increasingly promise it: nothing here generates a turnaround, an expression variant or a pose from a single source image. There are no AI features anywhere in the app. Every image on the sheet is one you or someone on the project actually drew, photographed or saved — the app's job is finding, rating and arranging that material, not producing new material to fill gaps in it.

Handing the sheet to someone else on the team

A sheet that only lives on the artist's own machine stops being useful the moment someone else needs to reference it — a colour supervisor checking palette consistency, an animator checking a turnaround angle, a director reviewing before a design gets locked. Because the canvas is a plain file sitting inside the project's own folder rather than something stored inside an app-managed library, handing it off is a file copy: drop the character's whole folder onto a shared drive, and the sheet opens on the next machine with every plate already in place, no export step and no re-linking of images that moved.

That also means a sheet doesn't silently go stale the way a flattened export does. A PNG export of a turnaround sheet is frozen the moment it's made — if the jaw shape changes in revision three, the export has to be redone from scratch. The canvas underneath keeps pointing at the actual reference files, so swapping one plate for an updated pose is an edit to the board, not a full re-export, and everyone working from the shared folder sees the same current version rather than a PNG someone forgot to replace.

When a design revision changes the reference midway through

Character work rarely locks on the first pass — a director asks for a different jaw shape, a colour supervisor pushes back on the palette, and half the sheet's source material is suddenly out of date. Because the sheet is built from a tag and a rating rather than a fixed export, updating it for a revision means re-running the same query and rebuilding the board from whichever references now have the highest rating, not starting over.

Old attempts that got superseded don't need deleting to keep the sheet accurate — lowering their rating below the floor the query uses is enough to drop them out of the set without losing the file, which matters when a director occasionally asks to see an earlier direction again before confirming the new one is actually better.

When a sheet needs to answer more than one question at once

Some sheets are genuinely doing two jobs — a turnaround for the animation team and a mood-and-palette reference for the colour team, built from overlapping but not identical material. Rather than building two separate canvases from scratch, a second board pulling the same character tag with a different section filter reuses the same tagged set without duplicating any files.

Both boards point at the same underlying files on disk, so a costume reference that gets rated higher after a revision updates its standing for whichever sheet queries it, rather than needing the change applied twice.

Common mistakes worth avoiding

Skipping the character tag and relying on folder names alone is the one that causes the most rework — a folder called Mira sketches doesn't survive being moved, renamed, or split across two project phases the way a tag on the files themselves does. Building the sheet before rating is another: laying out nine competing turnaround attempts and deciding between them on the canvas itself is slower than narrowing to four with a rating pass first, since the canvas is for arranging a decided set, not making the decision. And inventing a fresh one-off tag for every character-specific detail, instead of a note — a tag list that's mostly single-use tags stops being searchable, which defeats the reason to tag anything at all.

What to do when the cast grows past a handful

A single character's sheet is a small, self-contained board. A production with a dozen named characters and a hundred background designs is a different scale of the same problem, and the habits above hold up at that size without changing shape.

The character-name tag is what scales best — tag:any:mira-rook,castellan, harbor-crew returns every reference across a whole ensemble in one query, and organizing by project covers the pattern for keeping a production's full cast searchable as one workspace rather than a folder per character that stops being findable once there are more than a few. Keep the section vocabulary — turnaround, expression, palette — identical across every character; that's the part that turns "find every turnaround sheet across the whole cast" into a one-line query instead of a search through folders by hand.

For the workflow this sheet-building fits inside more broadly, reference management built for character designers goes further into visual development beyond a single sheet. If the sheets themselves need to be versioned as a project evolves — tracking how a design changed between drafts — version-controlling a moodboard covers doing that through the canvas file directly, since it's plain JSON a version control tool can diff. And for anyone weighing this against a tool built specifically for floating references over a drawing surface, the comparison with PureRef is worth reading — that's a different job from a searchable, taggable sheet, and the honest answer is that plenty of artists want both.

What to keep in mind before the next sheet

Two things carry the most weight across every sheet a production builds: the character tag applied consistently from the very first sketch, and a shared section vocabulary that never varies per character. Everything else above — rating, notes, reusing the canvas shape — is easier to get right once those two habits are in place from day one.

A sheet built this way is never really finished in the way a flattened export is — it stays exactly as current as the tagged, rated set behind it, which is the point, and which is worth trading for the small extra step of updating a query instead of clicking export. A production that revisits a character across three seasons doesn't accumulate three separate, drifting export files; it has one canvas that's been swapped and re-rated as the design evolved, with the folder's file history — or a version control tool pointed at the same folder — as the record of what changed and when, since the canvas itself is a plain file like any other.

Or try it against a character you're already working on — thirty days, every feature, no card and no account, enough to build one real sheet and see whether the tag-by-character habit is worth keeping across the rest of the cast.