Prompting Stickers, Emotes, and Chat Badges

Standard image prompts are calibrated for images that will be viewed at hundreds or thousands of pixels wide. Stickers, emotes, and chat badges are consumed at 28 to 112 pixels — a size at which subtle facial expression, fine texture, and delicate line work simply disappear. The model does not know your target size unless you tell it. When you prompt without size context, it generates at its native resolution with detail levels appropriate for large viewing. The result is an image that looks fine in preview and becomes a muddy, unreadable blob once scaled down to emote dimensions. The fix is to write prompts that demand bold, high-contrast, deliberately simplified designs from the start. Describe shapes as though you are explaining them to a sign painter working at arm's length: bold outlines, high contrast fills, no subtle gradients, exaggerated expressions. Think of every emote as a traffic sign — it must communicate instantly and unambiguously at the smallest size someone could possibly encounter it. Pile on the constraint language: 'thick black outline, solid flat fills, high contrast color palette, legible at 28x28 pixels, expressive and exaggerated, cartoon style.'

The visual grammar of sticker and emote illustration rests on three pillars: line weight, fill approach, and color budget. Line weight defines the outer and inner structure of the design. For emotes that must read small, prompt for 'thick black outline, 3-4px stroke weight equivalent, clean vector-style edges.' For stickers intended for large-format digital use like messaging apps, you can go slightly finer — 'bold outline, consistent 2px stroke, no weight variation' — but never as fine as editorial illustration. Fill approach determines how color is applied inside those lines. Flat fills without shading keep designs clean at small sizes: 'solid flat color fills, cel-shading style, no gradients, limited to 4 colors plus black and white.' Avoid painterly or airbrushed styles entirely; they introduce mid-tone complexity that destroys small-scale legibility. Color budget is your final constraint. Four to six colors maximum — including the outline and the background (which should be transparent) — is the professional standard for emote and sticker work. Specify each color explicitly: 'palette of coral red, warm yellow, soft cream, and black outline only, transparent background.' This precision prevents the model from inventing additional colors that break consistency across a set.

Emotes communicate emotion in a context where there is no time to read and no room for ambiguity. The emotion must be instantly legible to someone glancing at a chat stream. This requires what designers call expression coding — every element of the face is exaggerated in service of a single, clear emotional signal. Prompt for specific physiological markers of each emotion rather than naming the emotion abstractly. Instead of 'happy,' write: 'wide open smile, cheeks raised, eyes crinkled into upward arcs, eyebrows lifted, slight blush circles on cheeks.' Instead of 'angry,' write: 'heavy downward-angled eyebrows forming a deep V, pursed tight lips, furrowed brow lines, jaw set wide.' Instead of 'crying laughing,' write: 'mouth open in wide laugh, teeth showing, tears streaming from tightly squinted eyes, cheeks pushed up high.' The model responds better to anatomical description than to emotion labels because the latter are ambiguous — 'surprised' could mean forty different facial configurations. Pairing the anatomical description with a cartoon style amplifier — 'exaggerated, cartoonish, chibi proportions, large eyes relative to face size' — pushes the expression further toward the legible end of the scale.

The single most common production failure in AI-generated emotes and stickers is the non-transparent background. AI image models default to generating on a white or colored background. When you place a white-background emote on a dark chat overlay or a colored badge frame, the result is immediately broken. Prompting for transparency alone is inconsistent — models often render a white fill regardless. The reliable production workflow is to prompt for a white-background output with maximum edge contrast ('subject isolated on solid white background, clean edges, no drop shadows, no gradients near edges') and then route that output through a background removal step in Floniks' workflow editor. The combination of a clean edge prompt plus algorithmic background removal produces reliable transparent-background PNGs far more consistently than relying on the generation model alone. For badge artwork that lives within a circular or shaped frame, prompt the full design including the frame: 'circular badge design, thick colored border ring, centered character inside, designed to be cropped to circle, no content outside the circular boundary.' This tells the model to keep all meaningful content within the frame shape rather than bleeding important detail to the edges.

Professional emote and sticker packages ship as sets — typically five to twenty variants sharing a consistent art style, character design, and color palette. Generating them one at a time almost always produces stylistic drift: the fifth emote looks subtly different from the first because the model has no memory between individual generations. In Floniks, the solution is a batch workflow that carries the character and style spec as a fixed prefix through every generation in the set. Build a template node containing the character description — 'round-faced cartoon fox, cream chest fur, orange body, large dark eyes, thick black outline, 5-color palette: orange, cream, coral red, black, white' — and append it automatically to each emotion-specific description node. This prefix acts as a style anchor, dramatically reducing inter-emote drift compared to isolated prompts. Run all nodes in parallel within the workflow editor to minimize turnaround time. After generation, route all outputs through a shared background removal node and export to a consistent file naming convention. The result is a production-ready emote package delivered in a single workflow run rather than an hour of manual back-and-forth iteration.

Chat badges are displayed even smaller than emotes — subscriber tier badges on Twitch appear at 18x18 pixels — and they serve a different communicative purpose. Where emotes convey emotion, badges signal identity and status. They need to be instantly recognizable as belonging to a channel or brand, distinguishable from other badges in the same row, and readable in a single glance. Effective badge design prompts describe very simple iconic shapes: 'small circular badge design, centered white crown icon on deep purple background, thick border ring in gold, minimal detail, flat design, no text.' Avoid faces or complex scenes entirely at this scale; a single strong symbol is far more effective than any character illustration. When prompting a tiered badge set — where three badge variants represent increasing subscriber levels — the visual system should escalate clearly: 'three-tier badge set: Tier 1 in silver with single star, Tier 2 in gold with double star, Tier 3 in purple with triple star, all sharing the same circular format and border weight, legible at 18 pixels.' Describing the progression logic in the prompt lets the model produce a visually coherent hierarchy rather than three unrelated designs.