Creating a VTuber usually brings Live2D rigging, face tracking, and animation to mind. But before getting into any of that, you can first establish the character’s visual identity.
For this project, I used one original character (OC) across two PixAI Studio workflows to create a small VTuber starter kit:
- an 8-panel character profile
- a “Starting Soon” livestream screen
The main thing I wanted to test was character consistency. Could the same OC remain recognizable when moving from a simple character portrait into a much more complex streaming scene?
I used two existing PixAI Studio templates rather than building the workflows from scratch: Official Debut: Character ID Photos and Good Morning, 8,000 Viewers.
This focuses on VTuber visual assets. It does not cover Live2D rigging, face tracking, or building a real-time VTuber model.
1. Start With a Clearly Defined OC
My OC has a simple visual identity:
- long silver-lilac hair
- violet eyes
- black crescent-moon hair clip
- black and deep-purple outfit
- silver moon pendant
- purple, silver, and black color palette
I intentionally kept the design fairly simple.
When testing AI character consistency, having a few recognizable features makes it much easier to notice when the model starts drifting away from the original design.
Base OC prompt
A polished original anime VTuber character in a level, eye-level composition.
A young adult woman stands centered in the image. She has long silver-lilac hair with soft layered bangs, vivid violet eyes, and a small black crescent-moon hair clip on the left side of her hair.
She wears a modern black and deep-purple streamer outfit with subtle lavender accents, a fitted high-neck top, and a small silver crescent-moon pendant necklace. Her design is elegant and slightly futuristic without being overly complicated.
She has a subtle, confident expression and faces the viewer. Keep her face, hairstyle, hair accessory, outfit colors, and necklace clearly visible and recognizable.
Use a clean dark-purple gradient background that highlights the character without distracting from her design.
Polished anime character illustration, detailed face, clean linework, soft studio lighting, cohesive color palette, high visual quality.
Single character only. No text.
The character's purple-and-silver palette also gave me a visual direction I could reuse later in the streaming interface.
2. Generate VTuber Profile Images
For the first asset, I used PixAI Studio's Official Debut: Character ID Photos template.
Instead of rebuilding the workflow manually, I cloned the existing template and replaced its original character with my OC.
The workflow already included the structure for generating eight ID-style portraits arranged in two rows of four.
That meant I mainly needed to:
- Add my OC as the reference.
- Change the character prompt.
- Keep the existing profile layout.
- Generate the final set.
The template handled the presentation while I focused on preserving the OC's visual identity.
Result
The final set worked well as a profile image collection.
There were small differences between individual portraits, but the important features stayed recognizable: the silver-lilac hair, purple palette, and overall character identity.
This gave me a baseline before moving the character into a more complicated environment.
3. Create a Starting Soon Screen
Next, I cloned the Good Morning, 8,000 Viewers VTuber Streaming Scene template.
This workflow is considerably larger.
For this experiment, I only needed one part of it: the Starting Soon screen.
I did not continue into the later stream scenes, comments, video nodes, or other outputs.
Because this was a separate workspace, I uploaded the same original OC image again as the reference.
That choice was important.
I could have generated another character using the original text prompt, but even identical prompts can produce noticeable differences in facial structure, clothing, accessories, and hairstyle.
Using the same image reference gave both workflows a common starting point.
4. Adapt the Template Prompt to the OC
The original Starting Soon template was designed around another VTuber character.
Instead of replacing the entire prompt, I kept the useful layout instructions and changed the character-specific details.
For example, the template contained cat-girl styling that did not belong to my OC.
I replaced those details with:
- silver-lilac hair
- violet eyes
- crescent-moon accessory
- black and purple outfit
- celestial visual elements
Starting Soon prompt
16:9 horizontal Japanese VTuber YouTube livestream waiting screen. A polished anime VTuber "stream starting soon" waiting screen. This is a real VTuber pre-stream standby screen, not a generic livestream overlay.
Use my OC as the featured VTuber character. She is a young anime woman with long silver-lilac hair, violet eyes, and a small black crescent-moon hair clip. She wears a modern black and deep-purple streamer outfit with subtle lavender accents and a small silver moon pendant necklace. Keep her appearance recognizable and consistent with the original OC image.
Main center area: a clean and cute "Starting Soon" layout with a loading bar in the center. Japanese text: "配信準備中…" and "もうすぐ始まります". Show a loading progress bar around 60%.
Design style: soft pastel purple, lavender, pink, and silver color palette. Cute streamer aesthetic, elegant and slightly futuristic. Add small decorative details such as stars, moon motifs, soft UI panels, ribbons, and subtle lace-style accents. Make it feel cozy, polished, and suitable for a first VTuber stream.
Bottom left: a livestream chat window with only a few waiting comments. Top right: a LIVE badge and stream title.
High-quality anime VTuber waiting screen, clean composition, professional livestream overlay design, visually cohesive, cute and recognizable, single featured VTuber character, no extra characters.
The key change was that I did not limit the character identity to the character herself.
I also extended it into the interface.
Purple became the main UI color. The moon accessory became part of a wider celestial theme with stars and moon motifs.
That made the resulting screen feel more like something created specifically for the VTuber.
5. Compare the Outputs
The first generations did not all behave the same way.
Some preserved the character more accurately but switched to a vertical composition. Others followed the requested 16:9 layout better.
For a real Starting Soon screen, the horizontal composition made more sense, so I selected that version.
The final image retained most of the OC's important characteristics:
- silver-lilac hair
- violet eyes
- crescent-moon accessory
- purple outfit
- moon-inspired design language
The UI also adopted the purple and celestial theme, helping connect it visually to the character.
There were still a few small AI-generated text and UI imperfections, but the character remained recognizable.
6. Put the VTuber Starter Kit Together
I then placed the two final outputs together:
This was the most useful comparison.
The profile workflow focuses almost entirely on the character.
The Starting Soon screen introduces much more variation:
- a different pose
- background design
- interface elements
- stream text
- decorations
- lighting
- layout
Despite those differences, both outputs still looked like they belonged to the same VTuber.
The silver-lilac hair, violet eyes, purple-and-black clothing, and moon theme were visible across both designs.
I preferred the Starting Soon result because it made the character feel closer to an actual VTuber rather than simply an anime OC.
Character Consistency Matters More Than the Prompt
One useful lesson from this test was that a detailed prompt alone is not enough to maintain a character across multiple AI workflows.
A prompt can describe:
silver-lilac hair
violet eyes
crescent hair clip
purple outfit
But the model can still reinterpret those details.
The face may change.
The accessory may disappear.
The outfit may become completely different.
Using the same source image as a reference gives the model something more concrete to preserve.
It also helps to design the OC around a small number of visually distinctive traits rather than dozens of tiny details.
Where the Same OC Could Go Next
Once the visual identity is stable, the same reference could be reused for other VTuber assets:
- gaming overlays
- alternate Starting Soon screens
- Just Chatting layouts
- late-night stream scenes
- seasonal stream designs
- wallpapers
- social media graphics
The workflow would remain similar:
Original OC
↓
Reference Image
↓
Different PixAI Workflow
↓
Adapt Character Prompt
↓
Preserve Colors + Accessories + Theme
↓
New VTuber Asset
The goal is not simply to generate multiple images.
It is to make those images look like they belong to the same character and the same visual brand.
Conclusion
For this experiment, I turned one original character into a small VTuber starter kit using two existing PixAI Studio templates.
The workflow was essentially:
Original OC
↓
Official Debut
↓
8-Panel VTuber Profile
Original OC
↓
Good Morning, 8,000 Viewers
↓
Starting Soon Screen
Rather than rebuilding either workflow, I cloned the templates, replaced the original characters, adjusted the prompts, and reused the same OC reference.
The final results showed that it is possible to carry an OC's identity across different AI-generated VTuber assets before dealing with rigging or real-time animation.
For me, that is the useful part of this workflow: establish the character first, then build the streaming assets around that identity.
Top comments (0)