John Heine

Youtube e Empreendedorismo Digital

John Heine

Youtube e Empreendedorismo Digital

IA para Criadores

Como criar vídeos realistas com IA e personagem consistente

Criar uma imagem bonita com inteligência artificial já não é o maior desafio. O problema começa quando você tenta transformar aquele personagem em uma sequência de vídeos: o rosto muda, o cabelo ganha outro formato, a roupa troca sozinha e a identidade se perde entre uma cena e outra.

No vídeo que publiquei no YouTube, mostro o processo completo que usei na Higgsfield para criar Gabriel Reyes, montar uma ficha de consistência e gerar duas cenas cinematográficas de 15 segundos mantendo o mesmo personagem. O fluxo combina GPT Image 2.5 Flare, GPT Image 2.5 Sunburst e Seedance 2.5.

Acesse a Higgsfield pelo meu link: crie sua conta e confira o desconto de até 50% na primeira compra para novos usuários.

Este artigo contém um link de indicação. A oferta, os preços, os modelos e a quantidade de créditos podem mudar; confirme as condições exibidas pela plataforma antes da compra.

Criador produzindo versões consistentes de um personagem com inteligência artificial

O fluxo para criar um personagem consistente

O processo foi dividido em quatro etapas:

  1. criar o retrato-base do personagem no GPT Image 2.5 Flare;
  2. transformar o retrato em uma ficha de consistência no GPT Image 2.5 Sunburst;
  3. usar a ficha como referência no Seedance 2.5;
  4. manter personagem, roupa e direção de câmera, alterando apenas o cenário na segunda versão.

A ficha de consistência é a parte mais importante. Um prompt detalhado ajuda, mas o modelo de vídeo precisa de uma referência visual clara para preservar rosto, cabelo, barba, proporções e figurino durante o movimento da câmera.

Prompt 1: criação do retrato de Gabriel

Na Higgsfield, abra Image → Create Image, escolha GPT Image 2.5 Flare e configure a proporção vertical 3:4.

Professional photorealistic 3x4 ID portrait of Gabriel Reyes, a Latino man in his early-to-mid 30s, with warm medium-light olive skin, thick black naturally wavy hair with volume and loose strands falling over his forehead, deep brown almond-shaped eyes, thick dark eyebrows, straight proportional nose, defined oval face, natural jawline, short dark beard and connected mustache. He is looking directly at the camera, facing forward, head straight, shoulders aligned, calm neutral expression, lips closed, eyes open. Wearing a black denim jacket over a simple off-white crewneck shirt. No hat, no accessories. Pure solid white background (#FFFFFF), evenly distributed soft studio lighting, no shadows on the background, natural skin texture, realistic facial proportions, sharp focus, professional DSLR portrait photography. Centered head-and-shoulders composition, official ID photo framing, vertical 3:4 aspect ratio, high resolution. Preserve the exact facial identity from the provided reference photograph. No facial alterations, no beautification, no stylization, no CGI, no illustration, no artificial skin smoothing.

Prompt 2: ficha de consistência do personagem

Anexe o retrato aprovado, selecione GPT Image 2.5 Sunburst e use a proporção 16:9. A imagem deve conter corpo inteiro de frente, perfil lateral e um close grande do rosto.

Ficha visual com o mesmo personagem visto de frente, de lado e em close
Create an ultra-photorealistic CHARACTER CONSISTENCY REFERENCE SHEET of Gabriel Reyes.

Use the attached original photograph of Gabriel Reyes as the PRIMARY FACIAL IDENTITY REFERENCE. Use the attached layout example ONLY as a composition reference.

FORMAT:
Horizontal 16:9 aspect ratio, high resolution, pure white seamless studio background (#FFFFFF).

COMPOSITION — EXACTLY THREE VIEWS, SIDE BY SIDE:

LEFT (25% of image):
Full-body FRONT VIEW of Gabriel Reyes, standing naturally, facing directly toward the camera. Entire body visible from the top of his head to the soles of his shoes. Arms relaxed at his sides, neutral expression.

CENTER (25% of image):
Full-body SIDE PROFILE of the EXACT SAME Gabriel Reyes, standing naturally, facing right at a precise 90-degree angle. Entire body visible, including both shoes. Same height, physique, hairstyle and clothing as the frontal view.

RIGHT (50% of image):
Large, detailed HEAD-AND-SHOULDERS CLOSE-UP of Gabriel Reyes, facing the camera with a calm, neutral expression. His face must be clearly visible, showing his eyes, eyebrows, nose, lips, beard, hairline and realistic skin texture. This portrait is significantly larger than the two full-body figures.

CHARACTER IDENTITY:
Gabriel Reyes is a Latino man in his early-to-mid 30s, with warm medium-light olive skin, an elongated oval face, naturally defined jawline, deep brown almond-shaped eyes, thick black eyebrows, a straight proportional nose, medium lips and a natural short-to-medium dark beard connected to his mustache.

His hair is thick, black, medium-length, naturally wavy and slightly curly, with abundant volume and loose irregular strands falling over his forehead.

BODY AND CLOTHING:
Natural average-to-athletic male build, realistic body proportions. Black denim jacket over a simple off-white crewneck shirt, dark jeans and brown leather boots. Identical clothing in all three views. No hat, no accessories.

PHOTOGRAPHY:
Professional studio reference photography, bright and evenly diffused neutral lighting, subtle natural floor shadows, sharp details, realistic skin pores, natural anatomy, accurate proportions, neutral color grading.

LAYOUT REQUIREMENTS:
All three views appear in one single horizontal image, without separate boxes or panels. Clean continuous white background. Spacious, balanced composition. The two full-body figures remain completely visible. The facial close-up dominates the right side.

CRITICAL IDENTITY LOCK:
All three views must depict the EXACT SAME PERSON from the attached original reference photograph. Preserve identical facial proportions, nose shape, eye spacing, jawline, beard pattern, hairline, hair volume, skin tone and apparent age. Do not reinterpret or beautify Gabriel's face.

NEGATIVE CONSTRAINTS:
No additional people, no fourth view, no text, no labels, no borders, no collage decorations, no environment, no props, no facial changes, no inconsistent clothing, no illustration, no CGI, no artificial skin, no distorted anatomy.

FINAL RESULT:
A clean, realistic, professional 16:9 character reference sheet suitable for maintaining Gabriel Reyes' visual identity across AI-generated videos.

Prompt 3: vídeo cinematográfico no rancho

Em Video → Create Video, anexe a ficha, selecione Seedance 2.5 e configure 15 segundos, 16:9 e 720p. Mantenha a referência da imagem ativa no prompt quando ela aparecer como @Image1.

Create a 15-second ultra-photorealistic cinematic video in horizontal 16:9 format, using the attached Gabriel Reyes character consistency sheet as the primary identity reference.

CHARACTER:
Gabriel Reyes is a Latino man in his early-to-mid 30s, with thick naturally wavy black hair, deep brown eyes, a short natural dark beard, warm olive skin and the exact facial proportions shown in the reference sheet.

He wears his signature black denim jacket over an off-white crewneck shirt, dark jeans and no hat. His appearance must remain completely consistent throughout the shot.

LOCATION:
A simple, humble, authentic rural ranch in Mexico during golden hour.

A small, weathered adobe farmhouse with faded earth-toned walls, an old clay-tile roof, rough wooden beams, a modest covered porch and a dusty dirt courtyard.

A few simple wooden fences, sparse vegetation, dry countryside and distant hills complete the background.

The ranch looks genuinely lived-in, modest and peaceful, with subtle signs of age and everyday rural life. No luxury, no expensive decorations, no grand hacienda, no polished architecture.

Gabriel sits on a rustic wooden chair in front of the humble farmhouse, holding a natural wooden acoustic guitar.

CAMERA DIRECTION:
ONE SINGLE CONTINUOUS CINEMATIC SHOT.

The camera begins with a WIDE ESTABLISHING SHOT, showing Gabriel seated within the humble ranch environment. He is initially relatively small in the frame, surrounded by the old farmhouse, dirt courtyard and quiet Mexican countryside.

The camera then performs a very smooth, slow, physically realistic DOLLY-IN toward Gabriel, gradually transitioning from a wide shot to a medium shot and finally to an intimate close-up of his face.

No cuts, no transitions, no sudden zooms, no camera shake.

PRECISE TIMELINE:

0–4 SECONDS:
WIDE ESTABLISHING SHOT.

Gabriel is seated outside the humble adobe farmhouse, looking down at his acoustic guitar.

He gently strums one warm, resonant guitar chord.

The camera begins its slow forward movement toward him.

Show the modest ranch setting clearly while keeping Gabriel recognizable.

4–6 SECONDS:
MEDIUM SHOT TRANSITIONING INTO CLOSE-UP.

The camera continues moving closer.

Gabriel slowly lifts his head from the guitar and looks directly into the camera.

His expression is calm, sincere, compassionate and quietly hopeful.

By the end of this interval, the camera reaches a beautiful close-up of Gabriel's face.

6–11 SECONDS:
INTIMATE FACIAL CLOSE-UP.

The camera movement becomes extremely subtle, almost imperceptible.

Gabriel maintains natural eye contact with the lens.

With a warm, deep, slightly raspy male voice, he speaks clearly in natural Mexican Spanish:

"No te desanimes... Dios tiene algo más grande para ti."

His mouth movements must synchronize accurately with the Spanish dialogue.

His delivery is intimate and sincere, as if speaking directly to one person who needs encouragement.

No exaggerated facial expressions, no theatrical gestures.

11–15 SECONDS:
CLOSE-UP EMOTIONAL ENDING.

Gabriel finishes speaking.

He remains silent, maintains gentle eye contact and gives a small, natural, reassuring smile.

The camera holds the close-up, allowing the emotional message to resonate.

The warm sunset light softly illuminates his face.

A subtle rural atmosphere continues in the background.

LIGHTING:
Natural golden-hour sunlight, warm cinematic highlights, soft realistic shadows, authentic skin texture, shallow depth of field.

AUDIO:
One natural acoustic guitar chord at the beginning.
Clear Spanish dialogue.
Subtle countryside ambience, a gentle breeze and faint distant birds.
No background music.
No subtitles.
No additional voices.

VISUAL QUALITY:
Ultra-photorealistic live-action cinematography, professional lens rendering, natural human movement, realistic hands and guitar anatomy, subtle film grain.

IDENTITY CONSISTENCY:
Preserve Gabriel's exact facial identity from the attached reference sheet throughout the entire camera movement, especially during the transition from wide shot to facial close-up.

No face morphing, no changes in hairstyle, beard, age, clothing or facial structure.

ENVIRONMENTAL CONSISTENCY:
Keep the same humble Mexican ranch throughout the entire shot. No sudden background changes, no modern urban buildings, no luxurious hacienda, no expensive furniture.

FINAL PRIORITY:
The camera must start WIDE, move progressively CLOSER, and reach a FACIAL CLOSE-UP BEFORE Gabriel begins speaking. The dialogue must happen in close-up, not during the wide shot.

Prompt 4: mesma cena em uma cidade mexicana

Na segunda versão, personagem, roupa, fala, duração e movimento de câmera permanecem iguais. A mudança acontece apenas no ambiente. Essa comparação ajuda a verificar se a referência está preservando a identidade.

Mesmo personagem criado em um rancho e em uma cidade mexicana
Create a 15-second ultra-photorealistic cinematic video in horizontal 16:9 format, using the attached Gabriel Reyes character consistency sheet as the primary identity reference.

CHARACTER:
Gabriel Reyes is a Latino man in his early-to-mid 30s, with thick naturally wavy black hair, deep brown eyes, a short natural dark beard, warm olive skin and the exact facial proportions shown in the reference sheet.

He wears his signature black denim jacket over an off-white crewneck shirt, dark jeans and no hat. His appearance must remain completely consistent throughout the shot.

LOCATION:
An authentic urban city center in Mexico during golden hour. A beautiful pedestrian plaza with historic Mexican architecture, colorful building facades, stone pavement, a few distant pedestrians and warm late-afternoon sunlight.

Gabriel sits on a rustic wooden chair in the plaza, holding a natural wooden acoustic guitar.

CAMERA DIRECTION:
ONE SINGLE CONTINUOUS CINEMATIC SHOT.

The camera begins with a WIDE ESTABLISHING SHOT, showing Gabriel seated within the urban environment. He is initially relatively small in the frame, surrounded by the architecture and atmosphere of the Mexican city.

The camera then performs a very smooth, slow, physically realistic DOLLY-IN toward Gabriel, gradually transitioning from a wide shot to a medium shot and finally to an intimate close-up of his face.

No cuts, no transitions, no sudden zooms, no camera shake.

PRECISE TIMELINE:

0–4 SECONDS:
WIDE ESTABLISHING SHOT.

Gabriel is seated in the plaza, looking down at his acoustic guitar.

He gently strums one warm, resonant guitar chord.

The camera begins its slow forward movement toward him.

Show the Mexican urban setting clearly while keeping Gabriel recognizable.

4–6 SECONDS:
MEDIUM SHOT TRANSITIONING INTO CLOSE-UP.

The camera continues moving closer.

Gabriel slowly lifts his head from the guitar and looks directly into the camera.

His expression is calm, sincere, compassionate and quietly hopeful.

By the end of this interval, the camera reaches a beautiful close-up of Gabriel's face.

6–11 SECONDS:
INTIMATE FACIAL CLOSE-UP.

The camera movement becomes extremely subtle, almost imperceptible.

Gabriel maintains natural eye contact with the lens.

With a warm, deep, slightly raspy male voice, he speaks clearly in natural Mexican Spanish:

"No te desanimes... Dios tiene algo más grande para ti."

His mouth movements must synchronize accurately with the Spanish dialogue.

His delivery is intimate and sincere, as if speaking directly to one person who needs encouragement.

No exaggerated facial expressions, no theatrical gestures.

11–15 SECONDS:
CLOSE-UP EMOTIONAL ENDING.

Gabriel finishes speaking.

He remains silent, maintains gentle eye contact and gives a small, natural, reassuring smile.

The camera holds the close-up, allowing the emotional message to resonate.

The warm sunset light softly illuminates his face.

A subtle urban atmosphere continues in the background.

LIGHTING:
Natural golden-hour sunlight, warm cinematic highlights, soft realistic shadows, authentic skin texture, shallow depth of field.

AUDIO:
One natural acoustic guitar chord at the beginning.
Clear Spanish dialogue.
Subtle distant urban ambience.
No background music.
No subtitles.
No additional voices.

VISUAL QUALITY:
Ultra-photorealistic live-action cinematography, professional lens rendering, natural human movement, realistic hands and guitar anatomy, subtle film grain.

IDENTITY CONSISTENCY:
Preserve Gabriel's exact facial identity from the attached reference sheet throughout the entire camera movement, especially during the transition from wide shot to facial close-up.

No face morphing, no changes in hairstyle, beard, age, clothing or facial structure.

FINAL PRIORITY:
The camera must start WIDE, move progressively CLOSER, and reach a FACIAL CLOSE-UP BEFORE Gabriel begins speaking. The dialogue must happen in close-up, not during the wide shot.

Onde o Genjutsu entra no processo

Além do fluxo manual, a Higgsfield permite abrir o Genjutsu, acessar Community Trending e usar Recreate para adaptar uma tendência a um personagem próprio. No vídeo, mostro como adicionar Gabriel, conferir o custo em créditos e gerar uma versão em 720p.

Essa função acelera experimentos, mas não substitui direção criativa. Uma tendência pode fornecer estrutura e movimento; ainda cabe ao criador decidir se o resultado combina com o canal, a audiência e o posicionamento.

O que observar antes de gerar

  • use uma imagem-base frontal, nítida e com iluminação uniforme;
  • aprove primeiro o rosto antes de montar a ficha;
  • descreva identidade, figurino e restrições de maneira consistente;
  • indique o movimento da câmera em ordem temporal;
  • peça que o close seja alcançado antes da fala quando a sincronia labial for importante;
  • altere uma variável por vez ao comparar versões;
  • confira o custo em créditos antes de cada geração.

Vale a pena para canais dark e criadores?

A principal vantagem é centralizar modelos diferentes em um único fluxo. Isso pode ser útil para canais dark, narrativas com apresentadores virtuais, anúncios, videoclipes, séries curtas e projetos que precisam repetir o mesmo personagem.

Mas a ferramenta não elimina o trabalho de direção. A consistência depende da qualidade da referência, do prompt, das escolhas de enquadramento e da revisão do resultado. IA aumenta capacidade e velocidade; julgamento continua sendo humano.

Se você quiser testar o processo, acesse a Higgsfield pelo meu link de indicação e confira se a oferta de até 50% na primeira compra está disponível para sua conta.

NEWSLETTER DOS IMPARÁVEIS

Receba as próximas estratégias no seu e-mail

Novos artigos sobre YouTube, canais dark, monetização e IA útil. Conteúdo direto, sem spam.

JohnHeine

Empreendedor Digital desde 2011, criador de Diversos Projetos Online de Sucesso, Produtor do Curso Segredo Das Lives. Um dos Maiores Especialistas em Vídeo Marketing e Live Marketing do Brasil.

Deixe um comentário

O seu endereço de e-mail não será publicado. Campos obrigatórios são marcados com *