Character emotion and gesture library: production document
Why does a character need a separate document and not just art files
Without a fixed gesture library, a character 'floats': one illustrator makes the eyes larger, another changes the tilt of the head when surprised, and a third draws the hands shorter than the reference. In Japanese and Chinese fan art culture, as well as among VTuber models, there has long been a term 差分 (sai-bun) for a set of variations of one character. The Western industry calls a similar document an expression sheet or personbook — by analogy with a brandbook, only instead of a logo and colors, it contains facial expressions, poses, and image compression rules.
What is necessarily included in the document
A personbook is not reduced to an emotion gallery. The working structure includes several blocks, and skipping any of them creates a gap in production:
- Turnaround (reference list) — front, profile, 3/4 and back views, fixed on one proportion grid;
- Expression set — basic emotions with indication of which facial/body details change and which remain motionless;
- Gesture and pose library — characteristic hand movements, body tilts, interaction with objects and environment;
- Color samples and substrates — how a character is perceived against light, dark and brand backgrounds;
- Scaling rules — minimum and maximum size at which details remain distinguishable;
- File encoding — naming system to avoid manually searching for 'joy_profile_02' throughout the folder.
How many emotions to document: 9, 28 or 180+
There is no single correct number here — the set depends on where the character will live. The standard 3×3 grid is widely used for chat stickers: joy, laughter, sadness, tears, anger, surprise, love, embarrassment, contemplation — nine states that are sufficient for emojis and short correspondences. But this is not enough for animation or an interactive product.
| Size of the set | Where used | Features |
|---|---|---|
| 9 (grid 3×3) | Chat stickers, emojis, social networks | Minimal coverage, fast production, static PNGs |
| 28 full expressions | Comics, brand presentations, print products | Differentiating shades of one emotion: mild surprise / shock / fright |
| 180+ gestures (based on the Emotive Engine model) | AR, streaming, interactive applications, VTuber rigs | Categorization by body parts (hands, feet, torso, face), linked with NLP-API for generating reactions 'on the fly' |
The difference is not only in quantity but also in the logic of cataloging. For stickers, a flat list is sufficient. For interactive products, gestures must be arranged by body parts and states — otherwise, a developer will not be able to combine them programmatically, for example, overlaying a hand wave on a smile.
Technical parameters, without which the document is useless
The artistic part — half of the work. The second half — technical tolerances, which are often overlooked:
- A unified grid of proportions (head height as the unit of measurement) for all perspectives;
- The direction of light and the angle of shadow fall — are fixed once and applied to all poses;
- File formats by purpose: vector (SVG/AI) for scalable printing, PNG with alpha channel for web and stickers, spritesheet or rig for animation;
- The minimum display size at which the eyes and mouth are readable;
- For web and AR scenarios — a set of 'living' states: idle, wave, point, look, blink, with duration binding to an audio track for lip syncing;
- Base color codes in HEX, not 'by eye'.
Case: how the lack of a document canceled the mascot rebranding
In one project to update a mascot for a coffee chain, three consecutive contractors — an illustrator, a motion studio, and a mobile application contractor — worked in parallel without a single character bible. Result: the illustrator's character smiled, raising the left eyebrow, the animator's character had both eyebrows synchronized, and in the application the eyes changed more significantly than the reference. At the presentation, three versions of the same state looked like three different creatures to the client. We had to halt the animation for a week, align all assets to a single proportion grid, and retrospectively write a document that should have appeared at the start. After that incident, the rule in the team became simple: the character bible with turnaround, expression set, and file encoding is approved before starting any drawing, not after the first complaints.
Typical mistakes and how to avoid them
Most failures repeat from project to project, and almost all of them are preventable with simple rules at the start:
- Different scales of details between expressions. Solution — a rigid proportion grid and modular head height as a unit of measurement.
- Lack of negative states. Joy and surprise are drawn willingly, whereas sadness, fatigue, or irritation are done hastily — they are checked less often, although used no less frequently.
- No file encoding. Without a naming system, the team spends hours searching for the right variant in an archive of hundreds of PNG files.
- Ignoring the background. A character that reads perfectly on white may 'blend in' with a brand's dark background — color overlays need to be tested immediately.
- Calculating only for one channel. A set of 9 emotions for stickers attempts to be directly transferred into an AR application—there are not enough transitional and 'living' states like blink and idle there.
The personbook with an emotion library is not bureaucracy for the sake of bureaucracy, but rather a safeguard against the situation where the brand character looks like different characters at various contact points. The earlier the document appears in the production process, the cheaper any subsequent revision will be.
Why does a character need a separate document and not just art files
Without a fixed gesture library, a character 'floats': one illustrator makes the eyes larger, another changes the tilt of the head when surprised, and a third draws the hands shorter than the reference. In Japanese and Chinese fan art culture, as well as among VTuber models, there has long been a term 差分 (sai-bun) for a set of variations of one character. The Western industry calls a similar document an expression sheet or personbook — by analogy with a brandbook, only instead of a logo and colors, it contains facial expressions, poses, and image compression rules.
What is necessarily included in the document
A personbook is not reduced to an emotion gallery. The working structure includes several blocks, and skipping any of them creates a gap in production:
- Turnaround (reference list) — front, profile, 3/4 and back views, fixed on one proportion grid;
- Expression set — basic emotions with indication of which facial/body details change and which remain motionless;
- Gesture and pose library — characteristic hand movements, body tilts, interaction with objects and environment;
- Color samples and substrates — how a character is perceived against light, dark and brand backgrounds;
- Scaling rules — minimum and maximum size at which details remain distinguishable;
- File encoding — naming system to avoid manually searching for 'joy_profile_02' throughout the folder.
How many emotions to document: 9, 28 or 180+
There is no single correct number here — the set depends on where the character will live. The standard 3×3 grid is widely used for chat stickers: joy, laughter, sadness, tears, anger, surprise, love, embarrassment, contemplation — nine states that are sufficient for emojis and short correspondences. But this is not enough for animation or an interactive product.
| Size of the set | Where used | Features |
|---|---|---|
| 9 (grid 3×3) | Chat stickers, emojis, social networks | Minimal coverage, fast production, static PNGs |
| 28 full expressions | Comics, brand presentations, print products | Differentiating shades of one emotion: mild surprise / shock / fright |
| 180+ gestures (based on the Emotive Engine model) | AR, streaming, interactive applications, VTuber rigs | Categorization by body parts (hands, feet, torso, face), linked with NLP-API for generating reactions 'on the fly' |
The difference is not only in quantity but also in the logic of cataloging. For stickers, a flat list is sufficient. For interactive products, gestures must be arranged by body parts and states — otherwise, a developer will not be able to combine them programmatically, for example, overlaying a hand wave on a smile.
Technical parameters, without which the document is useless
The artistic part — half of the work. The second half — technical tolerances, which are often overlooked:
- A unified grid of proportions (head height as the unit of measurement) for all perspectives;
- The direction of light and the angle of shadow fall — are fixed once and applied to all poses;
- File formats by purpose: vector (SVG/AI) for scalable printing, PNG with alpha channel for web and stickers, spritesheet or rig for animation;
- The minimum display size at which the eyes and mouth are readable;
- For web and AR scenarios — a set of 'living' states: idle, wave, point, look, blink, with duration binding to an audio track for lip syncing;
- Base color codes in HEX, not 'by eye'.
Case: how the lack of a document canceled the mascot rebranding
In one project to update a mascot for a coffee chain, three consecutive contractors — an illustrator, a motion studio, and a mobile application contractor — worked in parallel without a single character bible. Result: the illustrator's character smiled, raising the left eyebrow, the animator's character had both eyebrows synchronized, and in the application the eyes changed more significantly than the reference. At the presentation, three versions of the same state looked like three different creatures to the client. We had to halt the animation for a week, align all assets to a single proportion grid, and retrospectively write a document that should have appeared at the start. After that incident, the rule in the team became simple: the character bible with turnaround, expression set, and file encoding is approved before starting any drawing, not after the first complaints.
Typical mistakes and how to avoid them
Most failures repeat from project to project, and almost all of them are preventable with simple rules at the start:
- Different scales of details between expressions. Solution — a rigid proportion grid and modular head height as a unit of measurement.
- Lack of negative states. Joy and surprise are drawn willingly, whereas sadness, fatigue, or irritation are done hastily — they are checked less often, although used no less frequently.
- No file encoding. Without a naming system, the team spends hours searching for the right variant in an archive of hundreds of PNG files.
- Ignoring the background. A character that reads perfectly on white may 'blend in' with a brand's dark background — color overlays need to be tested immediately.
- Calculating only for one channel. A set of 9 emotions for stickers attempts to be directly transferred into an AR application—there are not enough transitional and 'living' states like blink and idle there.
The personbook with an emotion library is not bureaucracy for the sake of bureaucracy, but rather a safeguard against the situation where the brand character looks like different characters at various contact points. The earlier the document appears in the production process, the cheaper any subsequent revision will be.
Recent articles
MegaFon, BMW, VTB and more 240+ companies. See how it looks in the showcase reel:
Our branding agency is approached to order:
Branding agency services
