FaceSwap: Replacing a Face or Head in an Existing Image
FaceSwap is one of ZenCreator’s core tools. Its job is to replace a face in an image — and in some modes, effectively transfer an entire head. This tool is especially important when working with models like SDXL or Flux, because those models often struggle to keep a consistent face across multiple images. FaceSwap is also useful with General models if you’re building your own pipelines or working on projects that require stable character identity.
Main FaceSwap Modes
1) Only Face
This mode replaces only the face oval area. That means:
hair usually stays the same,
the neck and the rest of the head are not affected,
only the facial mask changes (eyes, nose, lips, facial shape).
In the example shown in the video, the face changes, but the hairstyle remains the same, and even the neck stays untouched — that’s exactly how Only Face works.
2) Full Face and Full Face (Unusual Poses)
We recommend Full Face as the main mode in “about 90% of cases.” The reason is that it transfers not just the face, but almost the entire head. It would actually be more accurate to call it a “Head Swap,” because it affects:
the face,
hair / head outline,
the overall head area of the character.
2K vs 4K: What to Choose
In Full Face, you can choose 2K or 4K. Even though it’s tempting to always pick the highest quality, we generally recommend 2K, because:
it is more stable,
it produces fewer artifacts and unexpected shifts.
What to Know About Head Transfer
If the source image and the head-reference image differ a lot in hairstyle, hair volume, or head shape, you may see small changes in head position or alignment. In most cases, this is not a big issue, especially if your goal is consistency and quickly “fixing” identity.
Face Generator: Quickly Creating a Character Face
Next is Face Generator. It’s designed to help you quickly generate a high-quality character face that you can then:
use as your model’s “main” face reference,
apply in FaceSwap to replace faces,
plug into other workflows where recognizable identity matters.
In the examples, you get faces with strong detail, but they don’t look overly glossy or perfect — and that matters. When a face looks too perfect, it feels synthetic; slight imperfection makes it more believable.
How to Choose the “Main Face” for a Character
Faces can look similar at first, but still have meaningful differences. So it’s important to:
pick one face as the main reference,
ideally choose an image where the face is fully visible and straight,
then use it consistently in generations and face swaps.
You can also create a face using Text-to-Image by describing the portrait (age, ethnicity/type, eye color, hairstyle, etc.). The model will follow those parameters. You can even use expression-related prompt terms (for example, describing a specific “attitude” or facial expression) — in prompting, these are simply ways to define the vibe and facial expression.
Carousel: One Photo → Many Consistent Angles and Poses
Next is Carousel. In the examples, you can see results from a previously run task: the input was a few lingerie photos of a girl, and the output became a set of images that looks like one cohesive photoshoot:
the same character,
the same lingerie,
the same location,
repeating background details (like a lamp), which makes the series feel visually consistent.
How Carousel Works (Mechanics)
The workflow is simple:
Upload images (you can upload many — even “a thousand”).
Choose the aspect ratio.
If needed, use resize directly inside the tool.
Set how many variations you want as output.
This tool works without censorship, so details like tattoos and other character features can be preserved.
Collabs: Two-Character Photos (Collabs and Shared Scenes)
Next is Collabs. This tool generates photosets featuring two characters (for example, Instagram collabs or joint social scenes).
How it works:
upload a photo of the first character,
upload a photo of the second character,
optionally add a location reference,
or describe the location in the prompt.
Templates help you understand how to phrase requests — essentially: “the character from the first photo and the character from the second photo are doing X,” plus technical details.
Then you set:
aspect ratio,
pixel dimensions,
number of images.
The result is a consistent photoset with both characters, ready to use in content.
Lipsync: A “Talking Character” + Action Script During Speech
One of the key tools is Lipsync, which makes a character speak.
Important: at the time of recording, ZenCreator does not provide built-in options to:
choose a voice,
clone a specific person’s voice.
For higher-quality voice, you can use external tools. A common workflow is:
generate voice audio externally,
upload the audio and apply Lipsync in ZenCreator.
Why ZenCreator Lipsync Stands Out
This tool has two major strengths:
Works without censorship
If a scene gets blocked in other services, ZenCreator can be a reliable place to create a talking clip without restrictions.
You can describe actions during speech
This is a powerful feature: you can specify what happens while the character is speaking — even as a sequence:
the character walks,
after a few seconds starts speaking,
then stops,
then performs another action,
while the lip-sync remains accurate.
This makes it stronger than many alternatives, where you often get only a static “talking head” with no controllable scene dynamics.
Flows: Automating FaceSwap + Upscale (and Other Combo Processes)
Prebuilt flow scenarios combine multiple steps into a single run. These flows are “a bit outdated,” but still useful as quick automation.
Flow 1: FaceSwap + Upscale
The idea:
you provide a face image,
provide a body/scene reference,
choose the Upscale mode,
get an output where quality is improved (upscale) and then face swap is applied.
The benefit is speed: fewer manual steps.
Flow 2: Reference → Upscale → FaceSwap
A more advanced flow:
generates an image inspired by a reference,
applies upscale,
swaps in your model’s face.
The final image may differ in details (composition, jewelry, etc.), but remains clearly inspired by the reference — and the entire chain runs in one pass.
My Generations: Your Archive of Everything You’ve Created
My Generations is the library of all generations you’ve ever made. It’s useful when:
you have a large amount of content,
you need to find something from weeks or months ago,
you want to reuse content (send it into another tool or download it).
We recommend using this page as your main content archive and a fast search across your generation history.
Personas: Characters and Publishing Content
Next is an important part: ZenCreator isn’t only about generating — it also supports publishing.
On the Personas page, you can create “personas” that are linked to:
a face (currently the main supported reference),
later: body references and more,
name,
group (for organizing many personas),
bio (used later for caption generation and content plans),
connected social accounts (to publish content under that persona).
Why Use Personas If You Have FaceSwap and Local Files?
It’s mainly about organization:
you don’t store faces locally or search for them manually,
you select a persona, and the correct face loads automatically in tools like FaceSwap.
Personas also include:
Posts (where published posts will appear),
Gallery (content created using that persona’s face), which helps separate content by character instead of relying only on the general My Generations feed.
More references and analytics will be added here later.
Billing: Credits, Plans, Refunds for Errors, and Auto Top-Up
Finally, billing.
Where to Check Your Credits
in the profile header,
on the dedicated Billing page.
If you run out of credits, you’ll see a warning when trying to start a generation.
Refunds for Failed Generations
Important: if a generation fails with an error, the credits are refunded back to your balance. This reduces risk and makes experimenting safer.
Subscription Plans
At the time of the lesson, there are 4 plans:
Starter
Creator
Influencer
Professional
Plans will be updated soon, but currently Creator is the most popular option.
A new feature is coming soon: auto top-up. Users will be able to set a credit threshold, and when the balance drops below it, credits will refill automatically — so you can generate without constantly monitoring limits.
Mobile Access and the Telegram Mini App
ZenCreator is available:
in a desktop browser,
in a mobile browser (responsive UI — you can view and generate).
We also recommend Telegram: there is a ZenCreator mini app where you can log in and use the platform directly inside Telegram (via email/password or Google login).
That concludes the tools overview. The next block of training will focus not on the interface, but on how neural networks actually work and the core principles you need to understand to achieve stable, predictable results.