Choose Your AI Influencer's Look: Description or Images?

Quick Answer: Start From a Description or Images?

Start from a description when you have only a written character idea and want a photorealistic target. Start from images when the character already has an appearance and you are authorized to use those images.

When you start from a description, Generate mode attempts one composite reference sheet requested to show head-on, back, and right-side full-body views, then attempts photorealistic refinement. Those attributes are not validated or guaranteed; review the result.

A generated candidate is not guaranteed to be unique or exclusive and still requires the operator to review likeness and rights before use.

When you start from images, the current upload screen is labeled Reference Images (1-14) and displays a /14 uploaded counter. Those images should describe the same intended character rather than several competing versions.

If the target is non-photorealistic and you have no authorized images depicting the intended character in that treatment, neither current option is sufficient yet. Generate mode requests photorealistic refinement, while uploading needs existing visual inputs.

The simplest decision is:

• No established character image and a photorealistic target: start from a description.
• Established images you are authorized to use: start from images.
• Non-photorealistic target without authorized images depicting the intended character in that treatment: neither current option is sufficient yet.
• Unsure because your existing images disagree: do not upload the whole folder. Resolve the identity decision first.

This choice determines where the starting appearance comes from. It does not build the full character sheet, prove future consistency, or create the account strategy. After choosing, build the complete AI influencer character sheet and follow the full account-creation steps.

What You Already Have Determines How to Start

Both choices provide visual context for the character, but they begin with different assets.

Decision factor	Start from a description	Start from images
What you already have	A written character idea	Existing authorized images of the intended character
Starting input	Description, adult age, gender, and country context	Images added through the screen labeled Reference Images (1-14), which displays a /14 uploaded counter
Current visual treatment	Requests photorealistic refinement; review required	Defined by the supplied visual sources; downstream output is still not guaranteed
Initial control	Direct the concept with words, then review the result	Preserve more of an already chosen appearance
Preparation	Less file preparation	More responsibility for source quality and agreement
Initial generation	Attempts one composite sheet requested to show head-on, back, and right-side full-body views	Synthetic AI stores your selected images as reference images
Main tradeoff	The generated face may not match the exact person you imagined	Contradictory images can give Synthetic AI an unclear identity target

Neither choice is universally better. The useful choice is the one that matches what you have when setting up the character in Synthetic AI.

Start From a Description

Start from a description when you can describe the character but cannot point to an existing image that must define them.

In Synthetic AI, you provide a character description plus basic context. Generate mode attempts one composite reference sheet requested to show head-on, back, and right-side full-body views, then attempts photorealistic refinement. Those attributes are requested, not validated or guaranteed output properties; review the result before using it as the character's starting image.

Start from a description when:

• You want Synthetic AI to propose a new fictional identity from a description.
• The written concept matters more than matching an existing face exactly.
• You want one internally related set of angles rather than assembling source files.
• You are still open to reviewing the generated appearance and rejecting it if it does not fit the concept.
• Photorealism matches the visual direction you intend to pursue.

Starting from a description reduces preparation, not judgment. A detailed description can direct age, appearance, style, and outfit, but it cannot guarantee the exact face in your imagination. Review the output as a proposed identity, not as an automatic final answer.

Generating uses credits; check current pricing before you start.

What you gain

You begin with one generated composite candidate, not three separate files. That gives you one image to review when no earlier image defines the character.

What you give up

You give up exact control over the initial face. If the project already depends on a particular established appearance, repeatedly rewriting the description may be less direct than supplying the intended references.

Start From Images You Can Use

Start from images when the character's appearance has already been chosen before you enter Synthetic AI.

That could mean images you created yourself, commissioned sources you are authorized to use, or an established fictional character reference set under your control. Only upload images you are authorized to use and that depict the intended character.

The current upload screen is labeled Reference Images (1-14) and displays a /14 uploaded counter. The images added there become reference inputs associated with the character; they do not promise perfect reproduction in every later scene.

Start from images when:

• A specific existing appearance is non-negotiable.
• You already have a clear primary image of the intended character.
• You need Synthetic AI to see visual details that are awkward to reconstruct through text alone.
• Your source images agree on the face, age, hair, body, and other defining traits.
• You are prepared to exclude attractive images that depict a different version of the character.

The /14 uploaded counter is not a quality target. One clear source can be more useful than a large group of contradictory sources. After choosing images, use the character-sheet guide to design and test the complete reference pack.

What you gain

You keep Synthetic AI anchored to a visual identity that was selected before you started. This is the better choice when “make this established character” is more important than “invent a character from this description.”

What you give up

You take responsibility for preparing and selecting the inputs. The product cannot decide which of several conflicting faces is the real one for you.

Use This Decision Order

Answer these questions in order. Stop as soon as one choice becomes clear.

1. Is the intended starting treatment photorealistic?

• Yes: continue to question two.
• No, and you have authorized images depicting the intended character in that treatment: starting from images is the available option; continue to questions three and four before uploading anything.
• No, and you have no authorized images depicting the intended character in that treatment: neither current option is sufficient yet. Create or obtain authorized images first.

2. Is there already an image that must define the character?

• No: generate.
• Yes: continue to question three.

3. Are you authorized to use that image as an input?

• No or uncertain: do not upload it. If the target is photorealistic and you are willing to have Generate propose a different fictional identity, use Generate; otherwise resolve the input before uploading.
• Yes: continue to question four.

This is about choosing a usable input, not legal advice.

4. Do the available images depict one coherent identity?

• Yes: upload.
• No: reduce the set or settle the visual identity before uploading.

If the answers remain evenly split, the project probably has not decided whether it is inventing a new character or preserving an existing one. Resolve that question before generating more files.

Two Hypothetical Decisions

These examples illustrate the choice; they are not results from accounts operated by Synthetic AI.

A new home-organization character

The owner has a niche, an adult age range, a practical wardrobe direction, and a written personality, but no existing visual. No particular face needs to be preserved.

Decision: generate. The project needs a new proposed fictional starting identity, its target is photorealistic, and no earlier source needs to be preserved. The generated composite still requires review.

An established fictional fitness character

The owner already has several authorized images of the same fictional character. The face, hair, build, and styling direction are deliberate, and future content should begin from that identity.

Decision: upload. Recreating the character from text would discard useful source control that already exists.

The decision could change if those images turn out to depict conflicting versions. In that case, the next action is selection—not uploading all of them and hoping the model chooses correctly.

What This Choice Does Not Decide

Starting from a description or images does not answer:

• Which niche the account should enter.
• What the character should say or publish.
• Which views belong in a finished character sheet.
• How many posts should be created before launch.
• Whether the identity will stay consistent in every scene.
• How the account should grow or make money.

These questions come later. Follow the full AI influencer account build, then use the AI persona consistency steps when you are ready to control drift.

Keeping these jobs separate prevents a common failure: treating a good-looking image as though it were already a complete influencer business.

Common Description-or-Image Mistakes

Generating when an exact existing identity is required

A written description can guide a new character, but it is not a precise substitute for a visual source that the project must preserve.

Uploading every available image

The /14 uploaded counter is not a quality target. Conflicting sources create more ambiguity, not more control.

Expecting uploads to guarantee perfect identity

Uploaded images guide later generation. They do not remove the need for human review, rejection, and correction.

Using Generate for a non-photorealistic concept

The current Generate mode requests a multi-view composite and photorealistic refinement. Do not treat that as a guaranteed result or promise yourself a different treatment based on unsupported product behavior.

Treating authorization as a later problem

If you are not authorized to use a source image, do not upload it. If the target is photorealistic, use Generate to attempt a new proposed fictional reference image instead; otherwise resolve the input before uploading. Start with a usable input instead of trying to repair an input problem later.

FAQ

Can I start with only one AI influencer image?

Yes. The current upload screen is labeled Reference Images (1-14), displays a /14 uploaded counter, and allows you to continue after at least one image is added. This describes the current interface, not a claim that the backend strictly enforces a 14-image maximum. Whether one image gives enough useful coverage belongs to the full character-sheet process.

Does uploading images train a custom AI model?

No. Synthetic AI states that uploaded images and generated content are not used to train AI models. Uploaded files are decoded and re-encoded before storage and may be used as generation inputs, so the images serve as generation references; they do not train a custom model.

Does starting from a description create a complete character sheet?

No. It attempts one composite reference sheet requested to show head-on, back, and right-side full-body views, followed by an attempted photorealistic refinement. The requested views and finish are not validated or guaranteed. A complete character sheet adds selection rules, identity locks, useful views, versioning, and review standards beyond this first decision.

Are more uploaded images always better?

No. Additional images help only when they clarify the same intended identity. Images that disagree on defining traits can weaken the identity target.

Which option is easier for a beginner?

Starting from a description is usually simpler when the beginner has only a written idea and wants a photorealistic target. Starting from images is simpler when the beginner already controls a clear, coherent visual identity and does not need Synthetic AI to invent one.

Can either option guarantee a consistent AI influencer?

No. Both options establish starting visual context. Later instructions, scenes, image selection, model behavior, and human review still affect the output.

Make the Smallest Sufficient Choice

Start from a description when you want Synthetic AI to attempt a new proposed fictional composite reference sheet and the target is photorealistic. Start from images when you already control the identity you want Synthetic AI to follow.

Do not choose by counting features or files. Choose by answering one question: Is this project inventing the starting appearance, or preserving one that already exists?

Once that answer is clear, choose one option and move on to building the character.