Choose a canvas that fits the widest and tallest pose in your animation. A sword swing may need more room than an idle frame; tightly cropping each drawing can make the character jump when the frames play. Start with the intended size in the game, then choose one canvas and a stable anchor.
Keep three measurements separate:
Those measurements may be different. A character can occupy only 24 × 29 visible pixels inside a 32 × 32 frame, then appear much larger on screen because the game scales the sprite or camera.
Choose frame dimensions from the visual job the asset must do.
First decide how large the character or effect should look beside the rest of the game. Then identify the widest and tallest pose in the animation. A sword swing, squash frame, dust cloud, or hair movement may need more room than the idle pose.
Add only enough transparent room to contain those extremes and keep the intended anchor stable. Excess empty canvas wastes texture space; too little canvas clips motion or forces one frame to use a different size.
Common dimensions such as 16, 32, 48, or 64 pixels can be convenient for art direction, but they are conventions rather than requirements. Powers of two are not required for each individual frame. Your engine and target platform determine any limits that matter for the final texture.
Consistent width and height make a regular sprite sheet easy to slice. Consistent alignment makes it look stable.
Pick a logical anchor—often the point where a character's feet meet the ground, the center of a projectile, or the attachment point of an effect. Place that point at the same pixel coordinate on every frame canvas. The visible artwork can move around the anchor; the anchor itself should not drift accidentally.
Sprite Sheet Tool can accept mixed-size images. Its packer uses the largest source width and height as the cell, then centers smaller frames inside that cell. That is useful for a quick layout, but centering is not the same as preserving a foot, hand, or effect pivot. If exact alignment matters, normalize the source canvases in your art editor before packing them.
At 100% scale, a 32 × 48 frame and a 24 × 40 frame produce a 32 × 48 cell. The smaller image is placed four pixels right and four pixels down inside its cell. If a foot anchor sits at source coordinate (12, 39), it ends up at cell coordinate (16, 43). A foot at (16, 47) in the larger image therefore sits four pixels lower. Equal cells have not made those feet agree.
For a grounded character, put the intended foot point at the same coordinate on every source canvas before packing. Then compare extracted frames at native scale with a temporary crosshair at that point. The packer centers images; you need to align the feet yourself.
Also check the engine pivot. Unity's Sprite texture documentation describes how it positions a sprite. Set the pivot for the sliced sprites after aligning the source artwork; it affects placement, not which pixels belong to the frame.
For a gapless grid, the dimensions are straightforward:
sheet width = cell width × columns
sheet height = cell height × rows
rows = ceil(frame count ÷ columns)
For example, twelve 32 × 48 frames in four columns produce three rows and a 128 × 144 sheet when spacing and outer padding are both zero.
If you add spacing S between cells and outer padding P, use:
sheet width = (2 × P) + (cell width × columns) + (S × (columns - 1))
sheet height = (2 × P) + (cell height × rows) + (S × (rows - 1))
Write down the cell size, columns, rows, spacing, and padding. Your importer needs those same values. Guessing them later is a common cause of partial or combined frames.
Pack the prepared frames in the free sprite sheet editor, check the output dimensions, and preview the sequence. Keep the cell size with your asset notes. The image-to-sheet guide covers the export steps.
If you inherited a sheet, divide its usable width and height by the expected columns and rows. Subtract known margins and gaps first. A remainder usually means one of four things: the row or column count is wrong, the sheet includes padding, frames are not actually uniform, or the image was rescaled after export.
Use the sprite sheet splitter to test a row-and-column grid or an exact cell size. Inspect the outer cells and final row as well as the first frame; measurement errors often become clearer farther along the sheet.
A fixed cell is not mandatory for every rendering pipeline. A tightly packed texture atlas can store a different rectangle for each sprite, provided the runtime reads matching metadata and preserves the sprite's logical pivot separately. Large effects may also deserve their own atlas instead of forcing every character frame into an oversized cell.
Variable rectangles can save texture area, while pivots, timing and scale still need to be set in the game. Sprite Sheet Tool arranges the supplied pixels and reports their rectangles; it does not trim artwork or detect feet and pivots. Align the source art, then preview it before building animations around the export.

See how separate sprite frames become one image, how a game plays them, and when you need a regular grid or a JSON atlas.

A character reference sheet records a design. A sprite sheet contains game images. See how to turn an existing character into walking frames, or pack frames you already have.

Plan sprite animation FPS, loop duration, and per-frame holds with simple timing math, then carry those choices into your game engine.