Plan a VTuber avatar before commissioning the rig.

A single portrait is enough to discuss a visual direction. A PNG avatar workflow may need separately prepared still states. A Live2D character needs artwork that can be separated and rigged. A 3D avatar needs modeling and a suitable technical setup. These are different deliverables, not file extensions you can swap after generation.

Tell the artist how you plan to use the character: a face in a corner, a waist-up presenter, or a full-body avatar. Ask which source files and views they want. Their requirements should shape the concept brief before you spend time on costume details that will be outside the crop.

Lewdora does not create those rigs or integrations. Use it to explore a fictional adult character’s appearance, then take the approved concept into the appropriate production workflow. Keep that boundary clear when describing the project to collaborators.

Choose a face shape, hairstyle, and two or three dominant colors. View the concept at a small size against both light and dark backgrounds. Details that vanish at that size should not carry the entire identity.

Think about the eyes and mouth as working parts of the future avatar. Heavy hair across both eyes or a high collar across the mouth can complicate the expression brief. That does not make those designs impossible, but it gives the artist and rigger more problems to solve.

For an adult character, establish a mature identity through face, role, and wardrobe. Avoid child-coded styling. Lewdora is an adults-only service, and an adult age written in a prompt does not override an image that appears underage.

Proposed prompt, not a tested result: “Original fictional adult radio host in her thirties, mature face, chin-length dark teal hair tucked behind one ear, amber eyes, cream jacket over a dark high-neck shirt, small silver headphones resting around the neck, friendly neutral expression, waist-up anime character concept, simple background.”

Start with a neutral expression and an unobstructed face. Review hairstyle, eye shape, and the space around the mouth. Add a studio microphone later; an object crossing the face can make the concept less useful as an identity reference.

Keep the first accepted image as the appearance reference. If you change the costume, compare the face and hairstyle again. A successful outfit variation is not useful if it turns the host into a different person.

Choose a small initial set, such as neutral, pleased, concerned, and surprised. Name what changes in each expression: eyebrows, eyelids, and mouth shape. Keep the camera angle, hairstyle, outfit, and lighting stable while exploring them.

Compare the portraits side by side. Watch for changes to apparent age, jaw shape, eye spacing, or hairline. Where character reference is available, it can guide identity, but every candidate still needs review.

The artist may prefer to draw expressions from the approved base rather than repair separate generations. Ask before creating a large set. A clear written expression brief can save more time than dozens of inconsistent images.

Live2D’s official material-separation guide describes preparing separate image parts and PSD files for its workflow. Areas revealed by movement may need additional drawing. A flattened generated portrait does not contain those hidden areas as editable artwork.

Ask the artist and rigger which parts must be separated for your intended motion. Hair that crosses the face, dangling jewelry, and layered clothing all deserve discussion. The exact split depends on the design and rig, so a generic layer list is not a complete production specification.

Do not advertise a concept as “rig-ready” because it has a plain background. Background cleanup, part separation, redrawing, and rigging are separate tasks. If you are buying a finished avatar, ask for a written list of deliverables and supported uses.

Send one approved concept, the expression brief, intended crop, color notes, and identity features that must remain. Include a separate list of errors or uncertainties. A malformed headphone hinge should not become part of the character because it appeared in the generated image.

Disclose that you generated the concept with AI and ask whether the artist accepts AI references. Agree on how much they may redesign. Some artists will want to redraw the concept to make it consistent and practical to animate.

Use original designs and references you have permission to process. Avoid another creator’s face, costume, logo, or distinctive avatar as a shortcut to recognition. Check the terms for your generator, commissioned artwork, rig, and streaming destination before commercial use.

Separate image exploration from illustration, rigging, setup, and revisions in your project budget. Lewdora’s token price covers generation in Lewdora; it does not price the rest of an avatar project. There is no reliable conversion from a token count to a finished model cost.

Keep accepted concept files outside the browser gallery and record which version you approved. Once an artist starts, communicate changes as revisions to the brief instead of replacing the reference folder with new generations.

For a quick prototype, you may decide a manually prepared still image is sufficient. Describe it as a static avatar or concept. Upgrade the production format when the intended motion and budget are clear.