RetroChainer Shares Three-Week Guide to Building AI Influencers
RetroChainer publishes a guide on creating AI influencer accounts, covering niche selection, face generation with Nano Banana Pro and Pinterest references, and engagement tactics. The article frames the market as only months old and recommends sub-cultures, travel content and distinguishing features.
Original post · 8 min read
X ArticleI spent 3 weeks figuring out how to make AI influencers. Here's everything you need to know.
Step 1 - Choosing a Niche
The real secret is the niche. It determines everything:
The model's appearance and style
Content format and triggers
Audience engagement and account growth
The AI influencer market appeared literally a few months ago. The real potential of niches is still untapped copying successful girls makes no sense, you need to find your own direction.
Three working directions:
1. SubculturesAnime, cosplay, female streamers, football club fangirls. Timeless trends with maximally engaged audiences.
Examples: you can take a clip of a female CS2 streamer and replace her with your AI model. Or run an account of a cosplayer from the Marvel universe. A single post touching on a niche debate within a community can generate massive engagement just from one small detail.
2. Travel and eventsYour model can be on the Cannes red carpet, in Tokyo, or at a concert right now, without leaving home. This type of content feels expensive to produce because the viewer intuitively senses the effort behind it.
3. Physical featuresBirthmarks, scars, unique facial features as an additional trigger, but not as the foundation of the account. Unrealistic appearances caused a sensation a few months ago now it no longer surprises anyone.
Hypothesis: create a girl who is a fan of a football club and regularly mention one player as the best on the team. A wave of hate + a wave of support = bonus engagement and an algorithm boost.
Step 2 - Creating the Face
AI without references produces averaged-out looks technically attractive, but without character. You scroll through new AI accounts and every face looks copy-pasted. Here's why.
Tools:
Pinterest - searching for references
Nano Banana Pro (Higgsfield) - face merging
Process:
1. Find two photos of different girls with clearly visible faces
2. Upload both + the prompt to Nano Banana Pro:
Integrate a face into an existing scene. Substitute the face in the reference image with the face from the donor image. The objective is a seamless merge: the new face must inherit the exact expression, pose, and lighting interaction from the reference, while its color attributes (hair and eyes) are adapted from the donor for a perfectly harmonious and natural result.
3. Get the result. If needed add a distinguishing feature (birthmark, unusual eye color)
How to choose faces correctly
The main mistake is picking similar-looking faces. The neural network smooths out the differences and you end up with an averaged result again.
Pick contrasting faces and assign roles in advance:
First face base: sets the vibe of the niche (sharp cheekbones, "cold" look)
Second face donor: softens, adds attractiveness (baby face, full lips)
Face types by niche:
Cosplay, anime, game characters
Strong face goth, alt girl
Soft/romantic face fashion, luxury lifestyle
Hypothesis: find photos of girls of different ethnicities in the anime girl niche, merge them and add a birthmark as a distinguishing feature this instantly combines subculture and appearance triggers.
Step 3 - Creating the Video
The core of the method: AI replicates the movements of a person from the reference video and fully replaces the model in the footage.
3.1 Finding a Reference
Search on TikTok and Instagram. Use your niche as the search query goth girl, anime girl, cosplay.
Signs of a good reference:
High view count
Charismatic, expressive facial movements
Romantic undertone
Triggers from your niche
Key rule: the closer the person in the reference looks to your model — the more realistic the result. Kling struggles to transfer the movements of a long-haired girl onto a short-haired model.
3.2 Creating the First Frame
You can't just insert a photo of your model and hit "generate." Kling uses the background and pose from the uploaded photo, not from the reference. So first, you need to create a starting frame.
Upload to Nano Banana Pro:
1. A photo of your AI model
2. A screenshot of the first frame of the reference video
3. Prompt:
Take the girl's face and body from the first image, and the pose, emotion, and background from the second. Use the girl's face from the first image as the character's face and replace it in the second image, it is necessary to accurately convey emotion and playfulness, it is necessary to accurately convey the appearance of the girl from the first image without changing her appearance, the photo must be alive, the girl is not a doll, sincere real, photo taken on an iPhone phone camera.
3.3 Generating in Kling
Upload to Kling Motion Control: the first frame + the reference video. In the advanced settings, paste the prompt:
Use the attached reference video as the sole motion blueprint and transfer its movement onto the character from the attached photo(s), preserving the character's exact identity, body proportions, face and hair features, skin texture, clothing fit, and overall silhouette with zero morphing, zero style drift, and no added accessories; match the reference motion precisely frame-by-fr… continue on X ↗
The real secret is the niche. It determines everything:
The model's appearance and style
Content format and triggers
Audience engagement and account growth
The AI influencer market appeared literally a few months ago. The real potential of niches is still untapped copying successful girls makes no sense, you need to find your own direction.
Three working directions:
1. SubculturesAnime, cosplay, female streamers, football club fangirls. Timeless trends with maximally engaged audiences.
Examples: you can take a clip of a female CS2 streamer and replace her with your AI model. Or run an account of a cosplayer from the Marvel universe. A single post touching on a niche debate within a community can generate massive engagement just from one small detail.
2. Travel and eventsYour model can be on the Cannes red carpet, in Tokyo, or at a concert right now, without leaving home. This type of content feels expensive to produce because the viewer intuitively senses the effort behind it.
3. Physical featuresBirthmarks, scars, unique facial features as an additional trigger, but not as the foundation of the account. Unrealistic appearances caused a sensation a few months ago now it no longer surprises anyone.
Hypothesis: create a girl who is a fan of a football club and regularly mention one player as the best on the team. A wave of hate + a wave of support = bonus engagement and an algorithm boost.
Step 2 - Creating the Face
AI without references produces averaged-out looks technically attractive, but without character. You scroll through new AI accounts and every face looks copy-pasted. Here's why.
Tools:
Pinterest - searching for references
Nano Banana Pro (Higgsfield) - face merging
Process:
1. Find two photos of different girls with clearly visible faces
2. Upload both + the prompt to Nano Banana Pro:
Integrate a face into an existing scene. Substitute the face in the reference image with the face from the donor image. The objective is a seamless merge: the new face must inherit the exact expression, pose, and lighting interaction from the reference, while its color attributes (hair and eyes) are adapted from the donor for a perfectly harmonious and natural result.
3. Get the result. If needed add a distinguishing feature (birthmark, unusual eye color)
How to choose faces correctly
The main mistake is picking similar-looking faces. The neural network smooths out the differences and you end up with an averaged result again.
Pick contrasting faces and assign roles in advance:
First face base: sets the vibe of the niche (sharp cheekbones, "cold" look)
Second face donor: softens, adds attractiveness (baby face, full lips)
Face types by niche:
Cosplay, anime, game characters
Strong face goth, alt girl
Soft/romantic face fashion, luxury lifestyle
Hypothesis: find photos of girls of different ethnicities in the anime girl niche, merge them and add a birthmark as a distinguishing feature this instantly combines subculture and appearance triggers.
Step 3 - Creating the Video
The core of the method: AI replicates the movements of a person from the reference video and fully replaces the model in the footage.
3.1 Finding a Reference
Search on TikTok and Instagram. Use your niche as the search query goth girl, anime girl, cosplay.
Signs of a good reference:
High view count
Charismatic, expressive facial movements
Romantic undertone
Triggers from your niche
Key rule: the closer the person in the reference looks to your model — the more realistic the result. Kling struggles to transfer the movements of a long-haired girl onto a short-haired model.
3.2 Creating the First Frame
You can't just insert a photo of your model and hit "generate." Kling uses the background and pose from the uploaded photo, not from the reference. So first, you need to create a starting frame.
Upload to Nano Banana Pro:
1. A photo of your AI model
2. A screenshot of the first frame of the reference video
3. Prompt:
Take the girl's face and body from the first image, and the pose, emotion, and background from the second. Use the girl's face from the first image as the character's face and replace it in the second image, it is necessary to accurately convey emotion and playfulness, it is necessary to accurately convey the appearance of the girl from the first image without changing her appearance, the photo must be alive, the girl is not a doll, sincere real, photo taken on an iPhone phone camera.
3.3 Generating in Kling
Upload to Kling Motion Control: the first frame + the reference video. In the advanced settings, paste the prompt:
Use the attached reference video as the sole motion blueprint and transfer its movement onto the character from the attached photo(s), preserving the character's exact identity, body proportions, face and hair features, skin texture, clothing fit, and overall silhouette with zero morphing, zero style drift, and no added accessories; match the reference motion precisely frame-by-fr… continue on X ↗
♥ 755 · ⟲ 82 · 👁 1.6MView on X ↗
