My exact prompts and steps for making a cinematic AI scene of yourself, from character sheet to finished scene.
GPT Image 2Seedance 2.5Higgsfield
AI-generated still from the video
Before we begin
You're probably in one of three camps
1You're tired of showing up on cameraYou want content out without recording yourself every time.
2You want cinematic AI contentYou want scenes that look like film, not a talking head.
3You already have an AI clone or avatarYou're ready to put it into a real scene.
Either way, you're in the right place. In this guide I walk through how I made a cinematic AI scene with my clone: the references, the exact prompts and the settings.
My story About me
Why I'm building this in public
Nobody told me that quitting my job is the easy part. The hard part is everything after.
I spent my first year out unlearning how corporate taught me to operate, think and build. The failures were real too. Many times I wanted to quit. Many times I almost went back to a job. I had the drive. I didn't have the system. I was focusing on the wrong things the entire time.
That's why I built The Cloned Ecosystem. It starts with the brain: a Notion template that works as your content operating system. ClonedAI Studio is the creative studio, where you flip scripts, build prompts and generate video, audio, talking heads and cinematic scenes. Own The Funnel is the automation underneath: DM automations and a CRM, powered by Go High Level.
I needed the brain tight first. Now I'm building everything else on top of it, in real time. Watch, and learn from my mistakes. I also share the vulnerable, real parts of my life, so the connection with you is real.
Building in public. I am my own testimony. This is happening now.
What drained me
Before this I spent 11 years in insurance, taught yoga for 6 and was a lash tech for 2. I tried dropshipping and Amazon FBA, and I built and sold digital products.
Finding my voiceCreativity is hard, and it's easy to end up copying everyone else.
Hitting recordThe second the camera is on, you stop feeling like yourself.
Scripts and studio timeHand-written scripts before ChatGPT or Claude, plus paid studio time.
Hearing yourself backGetting comfortable on camera, and with your own voice on playback.
The warm marketFriends and family see it first. What will they think? Do I have anything valuable to say?
What drains you is rarely one thing. It's the whole process of making content.
Why an AI clone helped me
It shows up when I can'tMy content keeps going even when I'm burned out, so I can spend my time elsewhere.
Still work, not the draining kindI'm the creative person behind it. I direct.
Less pressureIt's still my voice and my brand, and still a little vulnerable. It just takes the pressure off.
From camera to directorA clone lets you do all of that without doing it all yourself. You become the director.
Why I built ClonedAI Studio
A different platform for everythingLearning every AI platform, tool, prompting method and skill was hard.
Burnout and information overloadI know the burnout and overload this causes.
Why people never startIt's probably too much to learn on top of your own career, especially if you run a business.
It feels too technicalFor a lot of people, it just feels like too much.
That's why I built it. So you don't need to be technical to get this done. It all lives in one place. Join the waitlist for early access.
Why now
Why avatars and clones, and why now
Because they solve a very real problem for people like me: people who have ideas, expertise, products, services and stories to share. People who know social media can create massive opportunities, and want to build a brand, sell digital products, promote services, grow influence or create content around their work.
But they don't want their real face, real energy and real schedule to be the bottleneck every single day.
You can now make a scene like the one at the top of this page without a camera, a studio or a crew.
1One photo, one consistent characterGPT Image 2 builds a character sheet from your face photo, so the same person shows up in every scene.
2Video models take it from thereSeedance 2.5 turns your sheet, a second reference and your voice clip into a scene.
3Less time in front of a cameraNo studio booking and no recording day for the scene.
Now and next. AI is for the now and for the future. It helped me, and it's helping a lot of other creators.
Before you start
What an AI avatar is (and isn't)
An AI avatar is not
A trick to fool your audience.
A gimmick for tech fans.
A toy you post once and forget.
A replacement for your business.
An AI avatar is
Made from a prompt: a new character, or one taken from a real photo.
A face and voice for your brand.
A way to show up without filming every day.
A channel that carries your message and expertise.
Labelled honestly as AI.
The way to think about itYour offer comes first. The avatar is the face that carries it.
AI clone vs AI avatar
Two ways to build a character. Know which one you're making.
AI-generated
AI clone. A digital version of you, trained on real images of yourself. Mine was made with Higgsfield Soul 2.0. It trains better with real images of yourself.
AI-generated
AI avatar. A character made from a prompt. Describe someone new, or extract the prompt from a real photo, then generate it. Mine was made with Nano Banana Pro.
How to make an avatar
Find a reference on PinterestPick an image of a person you'd like to use as the example for your avatar.
Extract the prompt in ChatGPTUpload the image and ask for a detailed text-to-image prompt for it.
Generate in Nano Banana ProPaste the prompt and hit Generate. You get a brand-new character with similar features to the reference.
Which one should you make?
Neither is better. It depends on your brand.
Choose an avatar if
You want creative freedom to design the character, the look and the niche.
You'd rather keep your real face off the internet.
You want several characters for several brands.
Choose a clone if
You already have a brand and people know your face.
You want the personal connection with your audience.
You'll mix real footage with AI content and want to look consistent.
Why I chose a clone. I already had a brand and wanted that personal connection. In this guide you start from a real photo of yourself, turn it into a character sheet with GPT Image 2, then use that sheet to make a cinematic scene in Seedance 2.5.
Step 01 GPT Image 2
Make your character sheet
Upload a front-facing photo of yourself, plus a body reference, and let GPT Image 2 build the sheet. Aspect ratio: 3:4, or let the model decide.
Your references
Your face photoFront-facing, head and shoulders, a very detailed face image. Aspect ratio: 3:4, or let the model decide.
Your body referenceUse pictures of your own body, from the neck down with no face. Or find a body type you like on Pinterest, crop out the face and use that.
Outfit, shoes and jewellerySearch Pinterest for everything you want your character to wear, or simply describe each item in detail in the prompt.
Describe an item like this: "The character wears sleek leopard-print luxury high-heel sandals inspired by Saint Laurent styling."
What the prompt gives you
AI-generated
Made with GPT Image 2 from the prompt below. Save it. You'll reuse it for every scene.
The full character sheet prompt
Upload order: face (@image1), body references (@image2 and @image3), outfit (@image4), shoes (@image5), jewellery (@image6), makeup (@image7 and @image8). Each one is a reference only. Swap my details for your own look.
Character sheet prompt
Ultra-realistic professional 5-panel fashion character reference sheet of the exact adult woman from @image1.
@image1 = SONIA FACE, IDENTITY AND PRIMARY LIKENESS REFERENCE ONLY. Preserve Sonia’s exact facial identity, skin tone, facial structure, proportions, and overall likeness.
@image2 and @image3 = BODY TYPE REFERENCE ONLY. Use these images only for Sonia’s body shape, proportions, curves, and overall silhouette. Do not copy the phone, mirror selfie composition, or plain activewear styling. Use them only to guide her physique: fit, toned, slim waist, curvy hips, full glutes, lifted medium-to-large chest, toned legs, feminine proportions.
@image4 = OUTFIT REFERENCE ONLY. Recreate the leopard-print dress from @image4 on Sonia. Do not copy a different model. Use only the garment design, cut, print, and silhouette.
@image5 = SHOE REFERENCE ONLY. Recreate the leopard-print YSL heels from @image5 on Sonia. Preserve the elegant ankle-strap high-heel sandal design and the distinctive sculptural gold heel detail. Use only as shoe reference.
@image6 = JEWELRY REFERENCE ONLY. Recreate the jewelry styling from @image6: layered diamond necklace, diamond bracelet, and glamorous statement rings. Keep the jewelry luxurious, polished, and elegant. @image7 and @image8 = MAKEUP REFERENCE ONLY. Use these only as glam makeup inspiration. Sonia’s makeup should be bold glam but not cakey, with smooth polished skin that still looks real and skin-like, not heavy or mask-like. SONIA HAIR:
Very long rich brown hair, dimensional brunette with subtle warm highlights and natural depth. Preserve Sonia’s brunette identity. Style it in a PROFESSIONAL BOUNCY BLOWOUT. Hair should be worn down, smooth and polished with soft body, lifted volume at the roots, glossy finish, and large rounded blowout bends through the lengths. The blowout should look luxe, salon-finished, feminine and glamorous, with soft face-framing movement. Not flat, not pin-straight, not beachy. Keep natural realism with individual strands, fine flyaways, and believable texture. Hair length should fall well past the chest.
SONIA FACE / MAKEUP:
Preserve Sonia’s exact face and South Asian / Punjabi identity. Warm medium-tan complexion with realistic skin texture. Wide elongated cat-eye eyes, naturally full dark brows, full lips, and softly defined cheek structure. Makeup should be glamorous and elevated, inspired by @image7 and @image8, but still refined and skin-like:
- flawless but realistic complexion
- bold lashes with separated volume
- softly sculpted cheeks and subtle contour
- softly glam neutral-brown eyeshadow
- defined eyeliner / lifted eye look
- softly highlighted skin
- nude to rose-nude full lips with a polished satin or soft gloss finish
- glamorous evening makeup, but NOT cakey, NOT over-powdered, NOT mask- like
Keep visible realism in the skin: subtle pores, soft texture, natural dimension, no plastic or airbrushed CGI look.
ACCESSORIES / STYLING:
Include the diamond jewelry from @image6:
- layered diamond necklace
- diamond bracelet
- statement rings
Optional small elegant earrings if they suit the look, but keep jewelry cohesive and luxurious.
Hands free. No handbag. No phone.
No sunglasses.
OUTFIT:
Recreate @image4:
- body-contouring leopard-print fitted midi/maxi dress
- structured, sculpted silhouette
- short sleeves / capped sleeve shape as shown
- defined waist
- sleek fitted skirt shape
- elegant, glamorous evening look
Preserve the leopard print pattern and overall silhouette faithfully. SHOES:
Recreate @image5:
- leopard-print Saint Laurent-style ankle-strap heels
- open toe
- elegant slim front strap
- high stiletto shape with sculptural gold heel detail
- glamorous luxury finish
BODY / PHYSIQUE:
Use Sonia’s identity with the body proportions inspired by @image2 and @image3:
- fit, toned, feminine physique
- slim waist
- curvy hips
- full glutes
- lifted medium-to-large chest
- toned arms and legs
- elegant, confident posture
Keep proportions realistic, flattering, and fashion-editorial while still human and believable.
5-PANEL PROFESSIONAL SHEET:
1. waist-up front
2. waist-up three-quarter
3. close-up beauty portrait
4. full-body front
5. full-body three-quarter
All five panels must show the SAME exact Sonia consistently:
- same face
- same body
- same long brown bouncy blowout
- same leopard dress
- same leopard heels
- same diamond jewelry
- same bold glam makeup
BACKGROUND / PRESENTATION:
Neutral warm-gray studio background.
Soft diffused professional studio lighting.
Polished luxury fashion reference-sheet look.
Clean panel arrangement.
No text.
No labels.
No CGI look.
REALISM / MATERIAL DETAIL:
Maximum realism:
- RAW skin realism
- realistic pores and natural skin texture
- individual hair strands
- believable hair shine and volume
- realistic dress fabric texture and seams
- realistic leopard print
- accurate jewelry sparkle and metal detail
- realistic shoe material and structure
- subtle optical softness and sensor grain
IMPORTANT NEGATIVES:
NO blonde hair
NO short hair
NO straight flat hair
NO ponytail
NO activewear
NO phone in hand
NO mirror selfie
NO sunglasses
NO handbag
NO cartoon look
NO artificial CGI look
NO overdone cakey makeup
NO heavy mask-like foundation
NO text
NO labels
NO duplicated character
What to keep: the five panels, the same-person list, the realism list and the negatives. Change the negatives to match your own look.
Step 02 Seedance 2.5
Create the scene
Take your full character sheet into Seedance 2.5, add the second person, and paste the scene prompt.
AI-generated
@image1 My character sheet from Step 01.
AI-generated
@image2 A clear photo of the second person (the friend).
Upload @image1Your character sheet from Step 01.
Upload @image2A clear photo of the second person.
Paste the promptCopy the full prompt below in one go.
The full scene prompt
Swap the names, looks, setting and spoken lines for your own scene. Keep the handheld rules, the timeline format and the pauses.
Scene prompt
For Seedance 2.5 Image to Video: EVERY SHOT IS HANDHELD.
Gentle constant shake, breathing sway, small drift and correction, slightly off-level horizons, faint motion blur at the frame edges.
Never a tripod, never a gimbal, never stabilized, never locked off, never a static frame.
Soft and slow, intimate, never jerky.
Smooth 60fps, natural real-life speed, no slow motion.
⚠️ SHOT PATTERN:
FRIEND, SONIA, SONIA, FRIEND, SONIA.
Coverage tightens each time it returns to a face, except the last shot which stays close.
SETTING
A condo balcony terrace at night during a small party.
SONIA and her female friend are both leaning on the balcony railing side by side, forearms on the rail, hips against it, turned slightly toward each other.
Close, shoulders almost touching, but not touching.
Below and beyond the rail:
a city at night, blurred streetlights, moving traffic far below, dark shapes of condo towers and buildings.
Behind them, out of focus:
the open balcony doors leading back into the party, dim interior light spilling through the glass, silhouettes of people walking past, occasional movement from guests inside.
A few people are farther down the balcony in deep background, completely out of focus.
Music bleeds faintly through the open doors from inside.
The balcony feels high above the city, quiet compared with the party behind them.
LIGHTING
Dim and directional.
Soft light from the condo doorway rakes across them from one side.
City lights below provide faint natural fill.
Everything else sits in deep shadow.
Deliberately underexposed.
Muted.
Low contrast.
Heavy organic grain.
No clean studio lighting.
No bloom.
No lens flare.
No HDR.
No crushed blacks.
No golden or amber cast.
Skin remains realistic and textured in the low light.
AUDIO
Muffled music from inside the condo.
Distant chatter.
Occasional laughter from the party.
Traffic far below.
Wind at height.
Their breathing is faintly audible during the quiet moments.
No cinematic score.
No emotional music cue.
The party continues indifferently behind them even when the conversation becomes serious.
Both women speak natural North American English.
THEM
FRIEND, the woman from @image2.
Same face, same identity and same physical appearance throughout.
She is dressed appropriately for an evening condo party.
Her appearance should feel polished but natural rather than editorial or overly glamorous.
Hair worn down and loose with natural movement from the balcony wind. Natural realistic skin texture remains visible.
Subtle evening makeup.
Simple delicate jewellery.
She begins the scene relaxed, curious and conversational.
She does not expect Sonia’s answer to become emotionally heavy.
SONIA, the woman from @image1.
Same face, same identity, same body proportions and same defining physical features throughout.
Long brown hair worn down with natural loose movement, realistic individual strands and subtle natural flyaways moving in the wind.
Keep Sonia’s natural grey/silver strands subtly visible through the brown hair, especially around the part and face-framing sections.
Realistic skin texture:
visible pores,
natural skin grain,
subtle uneven pigmentation,
under-eye texture,
fine lines,
natural highlights,
no beauty-filter smoothing.
Her makeup is polished but realistic:
elongated cat-eye,
defined separated mascara lashes,
soft neutral eye makeup,
full soft-pink glossy lips.
Simple elegant jewellery.
She is dressed for a sophisticated evening condo party in a fitted black evening outfit.
Nothing overly formal.
Nothing costume-like.
Her demeanor at the beginning is relaxed and comfortable.
When the conversation changes, she does not become theatrical.
The emotion gets quieter rather than bigger.
⚠️ THE TONE ARC
It starts casual.
Two women have stepped away from a party for a quieter conversation on the balcony.
They are relaxed, comfortable and talking softly underneath the noise coming from inside.
The friend asks Sonia what sounds like a simple relationship question. Sonia initially reacts like she expects to give an easy answer.
Then the question lands differently.
The entire emotional temperature changes without either woman raising her voice.
SONIA does not cry.
She does not become visibly dramatic.
She does not deliver a motivational speech.
The weight comes from how little she says.
Her answer should feel like something she has already lived through and accepted, but that still carries emotional history underneath it.
The final feeling should contain BOTH grief and peace.
Something deeply painful happened to her.
But she came out of it trusting herself more and feeling closer to God. TIMELINE
0.0 to 3.0s, FRIEND.
MEDIUM.
Handheld, framed chest-up.
She is leaning against the railing with the blurred city lights behind her.
Soft party light catches one side of her face.
She is turned slightly toward SONIA, forearms resting naturally on the railing.
Wind moves a few strands of hair.
She has a small conversational smile.
The question is not asked dramatically.
She thinks she is simply continuing an intimate late-night conversation between friends.
FRIEND, softly:
"What did your last relationship teach you?"
The small smile stays.
She genuinely expects a normal answer.
3.0 to 7.0s, SONIA.
MEDIUM.
Handheld, framed chest-up beside her friend.
SONIA is still leaning against the railing, city behind her.
She reacts at first with a tiny breath of amusement, almost like she is about to give an easy answer.
SONIA:
"Honestly?"
A tiny half-smile.
Then the question actually lands.
The smile disappears naturally.
SONIA:
"That's a big one."
⚠️ THEN NOTHING.
A full TWO AND A HALF SECONDS of real silence with the camera remaining on SONIA.
Do not cut away.
Do not shorten this silence.
Do not add dialogue.
SONIA’S eyes slowly move away from her friend and out over the city.
Her jaw tightens almost imperceptibly.
Her expression does not become sad in an obvious way.
She simply goes somewhere internally.
One slow breath out.
The party continues behind her.
7.0 to 12.0s, SONIA.
CLOSER.
Handheld.
The coverage tightens naturally.
Face and shoulders.
SONIA briefly looks back toward her friend and says the first part quietly.
No build.
No dramatic emphasis.
No inspirational delivery.
Like she is stating something she eventually had to accept.
SONIA:
"My gut knew the truth before I did…"
⚠️ A REAL PAUSE HERE.
TWO FULL SECONDS.
SONIA stops completely.
Her eyes leave her friend and move back out over the balcony railing toward the distant city.
The pause is essential.
Do not fill it with movement.
Do not cut away.
Do not shorten it.
Something in her face settles.
Not defeat exactly.
Recognition.
Then, still looking toward the city:
SONIA, quieter:
"…and losing him is what brought me back to God."
⚠️ She does NOT look back at her friend after saying it.
She continues staring out over the city.
No smile immediately after the line.
No tears.
No dramatic exhale.
Just stillness.
12.0 to 15.5s, FRIEND.
CLOSE.
Handheld.
Tight on her face.
Her original conversational smile is completely gone.
She does not react quickly.
Her eyes move slightly across SONIA’S profile, processing what she just heard.
Her lips part very slightly.
She takes a small breath.
Maybe the faintest swallow.
⚠️ SHE SAYS NOTHING.
Do not give her a reassuring line.
Do not make her touch Sonia.
Do not make her say “wow.”
Do not make her say “I’m sorry.”
She simply understands that the conversation went somewhere much deeper than she expected.
The party continues behind them.
15.5 to 18.0s, SONIA.
CLOSE, IN PROFILE.
Handheld.
Close on SONIA’S face, still turned outward toward the city rather than toward her friend.
Party light softly catches one side of her face.
City lights blur far below.
Wind moves a few strands of hair across her cheek.
She gives the smallest possible nod.
Barely visible.
As if acknowledging to herself:
yeah.
That is what it taught me.
Then a tiny, tired half-smile begins at one corner of her mouth.
Not happiness.
Not sadness.
Something closer to peace.
It does not fully reach her eyes.
She never turns back toward her friend.
Hold.
Cut to black.
VISUAL STYLE
EVERY SHOT HANDHELD with visible but gentle camera motion:
soft shake,
body sway,
small drift,
breathing movement,
tiny corrections,
slightly imperfect horizons.
No tripod.
No gimbal.
No stabilization.
No static framing.
Never jerky.
The camera feels operated by a real person standing close to them on the balcony.
Do not make the footage polished like a commercial.
It should feel observational and intimate.
⚠️ BOTH WOMEN REMAIN LEANING ON THE BALCONY RAILING FOR THE ENTIRE VIDEO. They remain side by side and close.
Neither woman stands upright.
Neither woman walks away.
Neither woman changes position dramatically.
Only subtle body movement:
breathing,
small weight shifts,
head turns,
eyes moving,
hair moving in the wind,
tiny gestures.
⚠️ THE TWO PAUSES ARE THE ENTIRE EMOTIONAL ENGINE OF THE SCENE.
PAUSE ONE:
The full two-and-a-half-second silence after SONIA says:
"That's a big one."
PAUSE TWO:
The full two-second silence after:
"My gut knew the truth before I did…"
Do not shorten either pause.
Do not insert additional dialogue.
Do not add reaction shots during these pauses.
Allow the uncomfortable silence to remain.
⚠️ SONIA LOOKS OUT INTO THE DISTANCE DURING THE SECOND HALF OF HER ANSWER AND NEVER LOOKS BACK AT HER FRIEND.
After:
"…and losing him is what brought me back to God."
nobody speaks again.
⚠️ THE LAST SIX SECONDS ARE ENTIRELY FACES AND SILENCE.
The friend processes what Sonia said.
Then Sonia remains in profile looking over the city.
No additional dialogue.
No hug.
No touch.
No comforting gesture.
No motivational conclusion.
The quiet is the conclusion.
STRICT PERFORMANCE NEGATIVE:
No crying.
No sobbing.
No trembling lips.
No dramatic acting.
No exaggerated sadness.
No motivational-speaker delivery.
No direct-to-camera performance.
No smiling through the emotional line.
No inspirational music.
No dramatic score.
No melodramatic reaction.
No hugging.
No touching.
No pacing.
No walking away.
No yelling.
No rapid dialogue.
No perfectly timed turn-taking.
No exaggerated facial expressions.
The emotion stays internal, restrained and psychologically real.
What to keep: the handheld rules, the timestamp format and the pause instructions. The silence is part of the shot.
Step 03 Higgsfield + your voice
Seedance 2.5 in Higgsfield
ClonedAI Studio runs its image and video generation on the Higgsfield API. Here's where everything goes.
Seedance 2.5 input screen in Higgsfield.
ReferencesTap + to add your character sheet (@image1) and your second person (@image2).
Voice clipAdd your audio clip under the images. The example is 0:30.
Prompt boxPaste your scene prompt. Tag each reference with Elements (@). Sound on.
SettingsModel: Seedance 2.5. 18s, 16:9, 480p. You can pick your own aspect ratio and resolution.
Use your own voice
Clone your voice (I use ElevenLabs) and upload it as the audio reference. Add your voice clip as @Audio1, then put this at the start of your scene prompt. Swap Sonia for your name. Listen to your clip first: it should sound conversational, not robotic.
Voice prompt
Use @Audio1 exclusively as the voice identity reference for Sonia. Match the speaker’s vocal texture, accent, natural pitch, pronunciation, and conversational cadence. The recording provides the voice reference only. Ignore its spoken words. Generate fresh speech using only the dialogue written below. Do not replay, quote, or include any dialogue from the reference recording.
Make it yours: this is the base. Add your own direction for the dialogue underneath it.
Checklist
Before you hit generate
Share yours
Recreate this cinematic scene and tag us
Tag us and use the hashtag so I can share your work, and you get a chance to win a free trial as a ClonedAI Studio beta tester.
@justsonia.ai@sonia.rayat#ClonedAIStudio
Meet ClonedAI Studio · exclusive first look
Shape the script. Create the scene. All in ClonedAI Studio
Replicate the winning formula from any viral video, turn it into your own original script, then generate a finished talking-head or cinematic video with your cloned voice. Your voice, images, video and talking heads in one place. Creators use a different platform for each. Here it's one.
Prompt Builder. Build a prompt, or upload an image to get its prompt.Script Flipper. Flip a viral script into your own. Powered by Claude.Generate Studio. Voice, image, video and HeyGen talking heads.Prompt Vault. My image and motion prompts, open to you.
Bring your own keys. ClonedAI Studio works with your own ElevenLabs and HeyGen keys.