Character Asset

The Character Asset node holds a reusable character: an approved portrait, a written description, sets of expressions, poses, angles, lighting variations and motion clips, and a voice. You build the character in the full-screen Character Studio, or choose one you already have. Wire the node into Generate Image or Generate Video, or mention it with @ in a prompt, and every generation shows the same person.

When to use it

• You want the same person in every scene of a story, an explainer or an ad.
• You need a cast of several characters that stay distinct across many shots.
• You want an expression library, a pose set or a turnaround sheet for one character.
• You want to reuse a character across workflows, or start from one in the Public Gallery.

Quick start

The node on the canvas

The node is a compact summary of the character. It shows:

• The approved portrait, the character's name, and its style and gender.
• A Role menu and an Identity lock menu. See Role and identity lock.
• The number of expressions, poses and motion clips.
• Whether a voice and a personality are set.
• A Studio button that opens the Character Studio.
• A Choose existing button, which becomes Replace once the node has a character. See Choose an existing character.

All editing happens in the Character Studio. The settings panel shows a summary, an Open Character Studio button, the Choose from library or gallery row and the Identity Lock menu.

Inputs and outputs

Role and identity lock

When the character reaches an image or a video model, its picture is described to the model as a short phrase, such as the person from reference image A. Two menus on the node set the defaults for that phrase.

Both defaults apply on image and video nodes, whether the character is mentioned with @, attached through the Assets input, or added as an extra reference, and also on runs from the API, the SDK and MCP. Each lock line names its reference image, so a scene with several characters stays clear.

A single mention can override the defaults. Pick another role on the mention's pill, or add ~lock or ~nolock to the mention, for example @maya:1:face~lock. Read Reference roles for every role and the full mention grammar.

The Character Studio

The Character Studio is a full-screen editor with a sidebar of pages in four groups. It saves as you work: a Saving… / Saved indicator shows the state, and there is no Save button.

Generation runs on the server. If you close the studio, the tab or the browser while an image or a clip is generating, the result is still added to the character, and its placeholder reappears when you open the studio again.

Profile

The Profile page holds the portrait and the core identity. It has the name, the description, the gender (male, female or other), the style (realistic, anime, 3d-pixar or illustration), the base outfit, the image model, and a seed prompt that frames the portrait.

Generate the portrait here. When a run makes several candidates, pick the one to keep and approve it. Approving a portrait also writes the character's canonical description, a detailed visual description that is sent with the character to later generations. Without it, the face drifts more between scenes.

References

The References page holds real photos of the person, up to seven, one per framing: front face, side left, side right, 3/4 left, 3/4 right, front body and other. The photos give the models more to match, so the generated pictures stay closer to the real person.

Pickers

The Pickers page holds two structured pickers whose choices are added to the character's own portrait and variation prompts automatically:

• Person: appearance attributes, such as hair, eyes, build, age, ethnicity, skin and face. It is the same catalog as the Person picker.
• Wardrobe: archetype, top, bottom, outerwear, footwear, headwear, accessories, color palette, material and era.

The pickers affect only this character's own portrait and variations. You do not wire them anywhere.

LoRA

Available on Nodaro Cloud

The LoRA page trains a character model from the character's own images. It is available on Nodaro Cloud only and is hidden on self-hosted installs.

1. The page shows every eligible image: the portrait, the reference photos, the expressions, the poses, the head and body angles and the lighting variations. All are selected. Clear the ones you do not want.
2. Keep at least 4 images selected. A counter shows how many are selected.
3. Click Train high-fidelity model. The button shows the price in credits, and training takes about 15 minutes.

When training succeeds, the page shows the model's trigger word. Re-train replaces the model with the images now selected, and Remove deletes it. Read Character training for how a trained character is used in generations.

Appearance

The Appearance page makes angle and lighting variations of the approved portrait, in three sections:

Expressions and poses

The Expressions and Poses pages make still images of the character from the approved portrait.

Each page has a generation bar with preset chips, a field for your own prompt, a model menu and a Generate button. Generate All queues every preset that is still missing and asks you to confirm first when it would start 4 or more generations. Every button and chip shows its credit cost for the chosen model.

The model menu offers Nano Banana Pro (default), Nano Banana 2, GPT Image 2, GPT Image 2.5 Flare, GPT Image 2.5 Sunburst and Seedream 4.5.

Motions

The Motions page makes short video clips of the character in motion, from the approved portrait. It needs an approved portrait first. The presets are walking, running, waving, sitting down, fighting stance, jumping, turning around, dancing and talking gesture.

The clips run on image-to-video models: Kling 2.6, Kling 2.5 Turbo Pro, Kling 3.0, Wan 2.6 I2V and Wan 2.7 I2V. A clip costs the same as the same image-to-video generation on that model. Motion clips have no refine action and no Generate All.

Actions on every card

Emotion videos

The Emotion videos page shows reference video clips of the character, grouped by emotion, when the character has any.

Sheet

The Sheet page composes a turnaround, expression or reference sheet from the character's images. See the Reference Sheet node.

Board

The Board page makes identity boards. An identity board is one dense 4K image that combines 2 to 12 of the character's own images, such as the portrait, expressions, poses, angles, lighting, wardrobe, close-ups, sheets and reference photos. A board is the strongest single-image reference for later generations.

• New board opens a picker with every image of the character, grouped by kind. Select 2 to 12 images. A number on each shows its place in the board. Generate makes the board at the 4K price of the Image Collage node.
• Duplicate reopens the picker with the same images selected, so you can try a variation. The copy gets a new name, such as "Evening gown 2", and never replaces the original.
• Rename a board by clicking its caption. Press Enter or click away to save, and Esc to cancel.
• Delete removes a board after you confirm.
• Cancel on a board that is generating stops it and refunds its credits.

A character can have up to 12 boards. In the @ mention menu, a character's boards are listed first, marked "board".

Voice

The Voice page gives the character a voice:

• Browse the premade voices, the public voice library and the custom voices you already have, with search and preview.
• Design a voice from a description, such as "a warm, gravelly old storyteller", and listen to it. A designed voice is for listening only and is not saved for reuse.
• Add voice traits, such as "deep, calm, British accent", to the chosen voice.
• Talk: type a line and click Speak to hear the character say it. With an approved portrait, Speak + lip-sync portrait also makes a short clip of the portrait saying the line.

Speak costs the same as Text to Speech. The lip-synced clip is priced per second of output, at the rate of the Lip Sync node.

Personality

The Personality page holds four notes about the character: Mood / Temperament, Speech Style, Movement Style and Behavioral Notes. They are saved with the character for your reference. Other nodes do not read them.

Borrow features from other assets

The Assets input of the Character Asset node takes features from other assets into this character's portrait. For each Character, Create Face, Location, Object or Animal/Creature node connected there, the node shows an Inject into portrait row with a menu of features:

Full likeness (default), Hair, Skin tone, Face, Style, Outfit or Personality.

For example, connect a first character and choose Hair to give a second character the same hair. The chosen feature is taken from the connected asset's description when the portrait generates. Text nodes and subject pickers connected to Assets add their words to the portrait prompt.

Choose an existing character

Instead of building a character, you can use one you already have. Click Choose existing on the node, or Choose from library or gallery in the settings panel. The picker has two tabs:

• My Library: the characters you saved.
• Public Gallery: characters shared with everyone. Choosing one copies it into your library first, and the node uses your copy. The Public Gallery is available on Nodaro Cloud and the Business edition.

Once the node has a character, the buttons read Replace, so you can swap in another character. The node takes the whole character: the portrait, every variation, the voice and the personality, so the nodes connected to it use the new character right away.

• If you pick a gallery character you copied before, the picker asks whether to use your copy or Make a new copy, so copies do not pile up.
• To remove a character from My Library, point at its card and click the trash icon. The character is archived and can be restored, and nodes that already use it keep working.

Read Community library for how shared characters work.

Credits

Generations in the studio are charged at the price of the model you choose, and every Generate button and preset chip shows that price.

• A motion clip costs the same as the same image-to-video generation on that model.
• A board costs the 4K price of Image Collage.
• Speak + lip-sync portrait is priced per second of output.
• Training a LoRA costs credits.

Tips

• Set Pickers and References before the first portrait. Their choices shape the very first portrait, so it lands closer to what you want.
• Approve the portrait first. Every expression, pose, angle and clip starts from it, and motion clips need it.
• Write a detailed description. Cover the face, the build, the hair and the clothing. The pickers cover the structured side.
• Compare before you replace. Use Add variation to keep the current card and try again beside it.
• Refine instead of rerolling. Use Refine to change one thing, such as "a wider smile", without starting over.
• Keep one style per project. Use the same style for every character of a project so they look like one world.
• Train when you have a spread of images. On Nodaro Cloud, train a LoRA from the cleanest on-model images, at least 4.

From the API

Every studio operation is also available to code and to AI assistants: create a character, generate and approve the portrait, add variations and motion clips, and archive or restore it. See Characters and the MCP tools.