Character creation
There are three reliable ways to create a character: build from scratch, build from a reference, or use yourself as the model. This lesson helps you choose the right path, lock the source setup, and approve one identity you can carry into later modification and video work.
You will leave withOne chosen creation path, one canonical source setup, one approved master identity, and three saved anchors.
Three ways to create the person
Start by choosing the source path. The prompt can be reused across tools, but the source path decides what the model is trying to protect.
From scratch
Use a structured prompt to invent a new creator identity when no face or reference exists yet.
- Best for
- Full control over audience fit, wardrobe lane, and realism.
- Inputs
- Creator brief + Structured prompt + Negative constraints
- Output
- A first master identity candidate.
Reference image
Start from one image and translate its face, styling, and mood into a reusable AI identity.
- Best for
- Speed, visual direction, and matching a look that already exists.
- Inputs
- Clean reference image + Identity notes + Structured prompt
- Output
- A generated identity that follows the source without copying it blindly.
Your own image
Use your own photo as the likeness anchor, then guide the model toward an influencer version of you.
- Best for
- Personal brands, founder-led content, and creators who want recognizable continuity.
- Inputs
- Your photo + Style reference + Structured prompt
- Output
- A creator version that still reads as you.
Generate and compare
Run the same structured prompt in two image tools, then compare the outputs before choosing the version that becomes your master identity.
Gemini
Gemini from scratch with the same prompt
This demo establishes the baseline: paste the complete character prompt into Gemini and generate a fresh person without using an existing face.
GPT image
ChatGPT from scratch with the same prompt
This demo uses the exact same prompt in ChatGPT, making it easier to compare realism, styling, and character fit without changing the instructions.
Show the prompt
{
"subject": {
"description": "A striking full-body vertical portrait of a young Indian woman with a rare combination of fair, light skin (Kashmiri/North Indian complexion) and piercing, natural green eyes. She is in her early 20s, standing casually in a light-filled apartment.",
"pose": "Full-body standing pose. She stands naturally, slightly shifted onto her left leg, with her right foot resting subtly forward. She is relaxed and confident; her body is oriented slightly towards the left (window light), but her head is turned to look directly at the camera. She is holding a small, worn paperback book in her left hand, which rests at her side; her right hand is casually tucked into her jeans pocket. She is standing on a patterned, handwoven Indian rug over a dark wooden floor.",
"expression": "A warm, relaxed, and natural resting face. She has a faint micro-smile, her lips softly closed. Her green eyes are looking directly and genuinely at the viewer; they possess subtle depth and natural brightness. Her expression is calm, approachable, and authentic.",
"details": {
"skin": "Very fair, light skin tone (North Indian complexion) with natural texture, visible pores, and faint imperfections. Light freckles are visible across her nose and cheeks. Faint under-eye coloration, subtle texture on the arms and hands, and slight dryness on her skin. Visible texture of peach fuzz on the face.",
"eyes": "Deep, natural green eyes with subtle limbal rings and variation in iris pattern. The sclera is a natural, healthy white with faint vein detail. A distinct, physically accurate catchlight reflection from the window on the left is visible in both eyes. The eyelashes are dark and natural, not overly long or dense.",
"teeth": "Not visible (lips are naturally closed in a faint micro-smile).",
"hair": "Dark, thick, wavy hair (typical Indian texture) with natural highlights and a soft, casual style. It falls past her shoulders, naturally layered and slightly messy from a day's wear. The hair has subtle shine and texture.",
"apparel": "She wears a casual yet stylish cream-colored cotton kurta (tunics) with subtle green and gold thread embroidery on the neckline and sleeves, and light wash, fitted denim jeans. The kurta has natural fabric creases and drapes authentically. She wears worn, white canvas sneakers on her feet.",
"accessories": "A thin gold chain around her neck with a small pendant; two small silver bangles on her left wrist; small hoop earrings. She wears a watch on her left wrist."
}
},
"environment": {
"location": "A cozy, sunlit apartment in New Delhi. She stands next to a large, traditional wooden window with natural daylight (late afternoon) streaming in from the left.",
"lighting": "Natural daylight, soft and ambient, primarily from the large window on the left. The light is soft and directional, creating subtle natural highlights on her skin and clothing. The right side of her figure has natural, soft shadows (Physically Accurate Lighting).",
"background": "A warm, lived-in apartment interior. Behind her, a wooden bookshelf filled with books, a potted Fiddle Leaf Fig plant, a blurred sofa with a decorative cushion, and traditional Indian wall art. The background is softly blurred (shallow depth of field), providing depth but keeping the focus entirely on the girl."
},
"camera": {
"type": "Vertical (9:16 aspect ratio), Full-frame DSLR (e.g., Sony A7RIII) with a prime lens.",
"lens": "35mm f/2.8",
"aperture": "f/4.0",
"depth_of_field": "Shallow natural DoF. She is fully in focus from head to toe, while the background (bookshelf, wall, furniture) is softly and realistically blurred (bokeh). The wooden floor and rug beneath her feet are also in the plane of focus, sharp near her shoes and softening further back.",
"focus_point": "Sharp focus on her green eyes; her face, hair, kurta, jeans, and sneakers are all in sharp detail. The blur begins naturally in the immediate background.",
"motion": "Static shot; the image is sharp and clear without blur from motion."
},
"image_quality": {
"resolution": "High resolution, capturing fine details without artificial sharpening.",
"grain": "Natural sensor grain (film-like quality), subtle and organic.",
"color_grading": "Neutral, natural colors. Slight warmth from the daylight. Skin tones are natural and true to the subject's complexion. No heavy filters or oversaturation.",
"imperfections": "Small specular highlights on the skin, natural fabric texture in the kurta and jeans, subtle creases in the clothing, realistic hair strands, minor scuffs on the sneakers. The wooden floor and rug have visible natural textures."
},
"style_constraints": {
"avoid": [
"Over-smoothed skin",
"Plastic or airbrushed texture",
"Perfect physical symmetry",
"Overly bright or white teeth (not visible)",
"Flat studio lighting",
"Heavy HDR or surreal look",
"Perfectly posed posture",
"High-contrast filter"
]
}
}Result image
Judge whether the face, skin texture, and wardrobe lane are stable enough to reuse.
Result image
Use this if it gives the cleaner identity, more natural texture, and stronger repeatability.
Task: copy the prompt, open GPT image generation, paste it, generate the image, and save the best output as your first identity candidate.
Understand the prompt controls
The prompt is not one long magic sentence. Each block controls a different part of the person, scene, and quality bar.
Face
Controls identity
Locks the features that must keep reading as the same person: eyes, skin texture, age signal, hair silhouette, expression, and overall face structure.Body and background
Controls pose and scene
Defines posture, body proportions, camera distance, apartment setting, daylight, and the physical believability of the frame.Costume
Controls styling
Sets the wardrobe lane, accessories, makeup level, and color direction without letting styling overpower the face.Quality constraints
Controls realism
Prevents waxy skin, fake symmetry, surreal light, extra fingers, plastic retouching, and other obvious AI failure modes.Video walkthrough
Edit the prompt, keep the structure
Watch the same structured prompt become a fresh persona by changing only the identity and styling blocks while the camera setup, body structure, and quality rules stay locked.
Generate the influencer from a reference image
The second path starts from one existing image. Use it to extract the identity cues you want, then rebuild that look into a reusable influencer version instead of copying the source blindly.
Reference image
One clean source photo where the face, styling, and lighting are easy to read.
Reference-image prompt
Use this prompt with the reference image when you want a new face identity while protecting the original body structure, silhouette, and proportions.
A realistic female model with a new identity and distinct facial features, meticulously maintaining the exact body structure and physical proportions of @element_1, in A minimalist, high-end fashion studio with a soft neutral-toned background, Centered full-body portrait showcasing the silhouette and posture, Full-length shot, Soft professional studio lighting with gentle highlights to emphasize the form. Hyper-realistic skin texture, natural hair detail, and premium fabric finish, Clean, professional lifestyle photography finish with sharp focus. Preserve @element_1 body structure and @element_1 physical proportions. Keep Realistic human appearance, Distinct face from the reference image, and Maintain silhouette of @element_1. Make the aspect ratio 9:16
Output images
Generated options that stay in the same visual family while giving you multiple identity candidates to review.
Use yourself as the model
The third path uses your own photo as the likeness anchor. Add a style reference and the same structured prompt to create an influencer version of yourself.
Your image
A clean photo where your face is readable and the lighting is not too harsh.
Sample generated image
A style reference that shows the visual direction you want the model to follow.
Prompt used
Use the same prompt structure, but replace the identity block with instructions that preserve your likeness.
Output image
Approve only if the result still feels recognizably connected to the original person.
Starter prompt scaffold
Use this for the from-scratch path when no face exists yet. It is intentionally shorter than the full reference JSON so you can control the identity before you add complexity.
Creator type: polished lifestyle influencer for ambitious young adults Audience signal: warm, aspirational, believable, not celebrity-coded Stable identity cues: expressive eyes, natural skin texture, repeatable hair silhouette, light makeup Canonical wardrobe lane: elevated casual neutrals with everyday premium styling Default environment: bright apartment daylight with realistic texture and soft DSLR depth Avoid: plastic skin, hyper-glam styling, surreal lighting, over-retouched beauty imagery
Reference prompt and image
Study one complete prompt-output pair before writing your own. This is the level of specificity we use when a character needs to stay believable under future modifications.
Reference prompt and output
This example shows the from-scratch path in practice. The prompt locks subject, pose, environment, camera, texture, and negative constraints so the result feels believable instead of plastic.
Notice how the from-scratch prompt moves from identity, to environment, to camera, to anti-drift constraints. That structure is what makes later modification easier to control.
subject
This block defines the human identity itself: description, pose, expression, skin, eyes, hair, clothing, and accessories. It is the main anti-drift layer.
Identity, pose, expression, and the stable visual traits that must keep reading as the same person.
environment
This block tells the model where the creator exists and what kind of light shapes the realism. It controls mood without replacing the identity.
A sunlit New Delhi apartment with soft late-afternoon window light.
camera
This block locks framing, lens feel, depth of field, focus, and stillness. It keeps the image photographic instead of floating into generic AI aesthetics.
Vertical 9:16 framing, DSLR realism, and sharp focus on the eyes.
image_quality
This block protects texture. It tells the model to preserve pores, fabric detail, natural grain, and believable color instead of over-polishing the output.
High detail, natural grain, realistic textures, no oversaturated finish.
style_constraints
This block prevents the most common failure modes. It is where you explicitly tell the model what kind of fake perfection to avoid.
No plastic skin, no heavy HDR, no flat studio light, no over-posed posture.
{
"subject": {
"description": "A striking full-body vertical portrait of a young Indian woman with a rare combination of fair, light skin (Kashmiri/North Indian complexion) and piercing, natural green eyes. She is in her early 20s, standing casually in a light-filled apartment.",
"pose": "Full-body standing pose. She stands naturally, slightly shifted onto her left leg, with her right foot resting subtly forward. She is relaxed and confident; her body is oriented slightly towards the left (window light), but her head is turned to look directly at the camera. She is holding a small, worn paperback book in her left hand, which rests at her side; her right hand is casually tucked into her jeans pocket. She is standing on a patterned, handwoven Indian rug over a dark wooden floor.",
"expression": "A warm, relaxed, and natural resting face. She has a faint micro-smile, her lips softly closed. Her green eyes are looking directly and genuinely at the viewer; they possess subtle depth and natural brightness. Her expression is calm, approachable, and authentic.",
"details": {
"skin": "Very fair, light skin tone (North Indian complexion) with natural texture, visible pores, and faint imperfections. Light freckles are visible across her nose and cheeks. Faint under-eye coloration, subtle texture on the arms and hands, and slight dryness on her skin. Visible texture of peach fuzz on the face.",
"eyes": "Deep, natural green eyes with subtle limbal rings and variation in iris pattern. The sclera is a natural, healthy white with faint vein detail. A distinct, physically accurate catchlight reflection from the window on the left is visible in both eyes. The eyelashes are dark and natural, not overly long or dense.",
"teeth": "Not visible (lips are naturally closed in a faint micro-smile).",
"hair": "Dark, thick, wavy hair (typical Indian texture) with natural highlights and a soft, casual style. It falls past her shoulders, naturally layered and slightly messy from a day's wear. The hair has subtle shine and texture.",
"apparel": "She wears a casual yet stylish cream-colored cotton kurta (tunics) with subtle green and gold thread embroidery on the neckline and sleeves, and light wash, fitted denim jeans. The kurta has natural fabric creases and drapes authentically. She wears worn, white canvas sneakers on her feet.",
"accessories": "A thin gold chain around her neck with a small pendant; two small silver bangles on her left wrist; small hoop earrings. She wears a watch on her left wrist."
}
},
"environment": {
"location": "A cozy, sunlit apartment in New Delhi. She stands next to a large, traditional wooden window with natural daylight (late afternoon) streaming in from the left.",
"lighting": "Natural daylight, soft and ambient, primarily from the large window on the left. The light is soft and directional, creating subtle natural highlights on her skin and clothing. The right side of her figure has natural, soft shadows (Physically Accurate Lighting).",
"background": "A warm, lived-in apartment interior. Behind her, a wooden bookshelf filled with books, a potted Fiddle Leaf Fig plant, a blurred sofa with a decorative cushion, and traditional Indian wall art. The background is softly blurred (shallow depth of field), providing depth but keeping the focus entirely on the girl."
},
"camera": {
"type": "Vertical (9:16 aspect ratio), Full-frame DSLR (e.g., Sony A7RIII) with a prime lens.",
"lens": "35mm f/2.8",
"aperture": "f/4.0",
"depth_of_field": "Shallow natural DoF. She is fully in focus from head to toe, while the background (bookshelf, wall, furniture) is softly and realistically blurred (bokeh). The wooden floor and rug beneath her feet are also in the plane of focus, sharp near her shoes and softening further back.",
"focus_point": "Sharp focus on her green eyes; her face, hair, kurta, jeans, and sneakers are all in sharp detail. The blur begins naturally in the immediate background.",
"motion": "Static shot; the image is sharp and clear without blur from motion."
},
"image_quality": {
"resolution": "High resolution, capturing fine details without artificial sharpening.",
"grain": "Natural sensor grain (film-like quality), subtle and organic.",
"color_grading": "Neutral, natural colors. Slight warmth from the daylight. Skin tones are natural and true to the subject's complexion. No heavy filters or oversaturation.",
"imperfections": "Small specular highlights on the skin, natural fabric texture in the kurta and jeans, subtle creases in the clothing, realistic hair strands, minor scuffs on the sneakers. The wooden floor and rug have visible natural textures."
},
"style_constraints": {
"avoid": [
"Over-smoothed skin",
"Plastic or airbrushed texture",
"Perfect physical symmetry",
"Overly bright or white teeth (not visible)",
"Flat studio lighting",
"Heavy HDR or surreal look",
"Perfectly posed posture",
"High-contrast filter"
]
}
}What you should finish with
This topic is complete when these outputs exist and are saved for the next stage of the workflow.
- One chosen character-creation path: scratch, reference, or self-model.
- One canonical source setup for the winning path.
- One approved master identity choice for future generations.
- Three saved anchors that prove the identity is stable enough to reuse.
Placeholders for uploads
These are the assets we will plug in later. Keeping the slots visible now makes the workflow feel complete and shows exactly what still needs to be collected.
placeholder
From-scratch winner set
Upload the first approved scratch batch and mark the frame that becomes the identity baseline.
placeholder
Reference image + generated match
Upload the source reference image beside the generated result that best preserves the identity cues.
placeholder
Self-model source set
Upload the approved self-images and the first generated version that still reads as you.
placeholder
Creation path scorecard
Upload the sheet used to compare scratch, reference, and self-model paths before locking the winner.
Once the identity is locked, Character modification is where you keep the same person while changing outfit, styling, and scene without breaking consistency.
Continue to Character modification