Tsubaki.2 Tutorial Challenge: Winners, Workflows & the Best Community Guides

The Tsubaki.2 Tutorial Challenge has officially wrapped up.

Over the course of the event, creators from across the PixAI community shared their techniques, workflows, and discoveries, covering everything from beginner-friendly prompting to advanced prompt engineering for anime AI image generation.

Thank you to everyone who participated. Every submission added something to the community’s collective understanding of how to use Tsubaki.2 well.

Below, you’ll find a curated overview of all submissions organized by topic, so you can quickly find the tutorials most relevant to your workflow. Whether you are new to PixAI, exploring Tsubaki.2 for the first time, or refining advanced prompt engineering techniques, there is something here for you.

Table of Contents

Browse by Topic

Whether you are looking for your first Tsubaki.2 prompt framework or trying to solve a specific anime image generation problem, start with the section that matches your goal.

Beginner Guides: Getting Started with Tsubaki.2

New to PixAI or AI image generation in general? These tutorials will get you generating in minutes.

なるみなみ — PixAIチュートリアル Tsubaki.2 (Tsubaki.2 Tutorial for PixAI)

The most hands-on onboarding guide in the batch. It walks through the full generation flow, from opening the interface to adding a LoRA, managing Credits, and publishing your first image, with screenshots at every step. A strong choice for anyone starting from zero.

なるみなみ— AIイラスト生成サイトPixAIの最新オリジナルモデル『Tsubaki.2』プロンプトの知識が無くてもオッケーよん！(You Don’t Need Prompt Knowledge for PixAI’s Latest Model Tsubaki.2!)

Starting from a single word — “street fashion” — this tutorial documents an entire iterative session in real time: adding focus, perspective, motion lines, and style presets step by step. The best demonstration of what Tsubaki.2’s natural language input actually looks like in practice.

@teshio_k — PixAIの新フラグシップモデルTsubaki2の衝撃を徹底解説 (A Complete Breakdown of PixAI’s New Flagship Model Tsubaki.2) · Video

A comprehensive video introduction to Tsubaki.2’s core strengths, including multi-character generation, lighting control, shadow handling, and structured prompting. It also compares how the same prompt behaves across different Style presets, making it one of the strongest all-in-one overviews for new users.

Tsubaki.2 Prompt Engineering: Structure, Logic, and Control

The tutorials that go deepest on how to write prompts that Tsubaki.2 actually responds to.

canglanxi — AI Image Generation: From Shopping List to Visual Design

One of the most rigorous frameworks in the collection. It argues that DiT-based models like Tsubaki.2 respond better to structure than to keyword stacking, and introduces a seven-step prompt order: Composition → Lighting → Depth of Field → Angle → Subject → Environment → Details. A particularly strong resource for users interested in prompt engineering rather than trial-and-error prompting.

Traditional Chinese version: AI生圖入門：從購物清單到設計畫面

青井透子｜日常の過剰解釈屋 — 【PixAI Tsubaki.2】“静かに狂う”を作るプロンプト設計｜映画的ライティングと違和感の作り方 (Prompt Design for “Quietly Unhinged”: Cinematic Lighting and Uncanny Effects in Tsubaki.2)

A three-stage workflow for atmosphere-driven image generation: start with character, layer in cinematography (lighting, focus, film grain), then add narrative meaning through a single detail (what’s reflected in the sunglasses). Every stage includes the actual prompt and a comparison image. The most transferable framework in the collection — works for any mood, not just this one.

@teshio_k —Tsubaki.2 徹底活用ガイド：次世代アニメ画像生成の技術と実践 (Tsubaki.2 Complete Guide: Techniques and Practice for Next-Gen Anime Image Generation)

A broad and practical reference covering three levels of prompt formulas, lighting and camera control, style blending, and recommended use cases for different generation modes. It also includes full multi-character prompt templates, making it useful for both intermediate and advanced users.

@mai_et_pizza —Mastering Tsubaki.2’s Logic

A compact but insight-dense thread covering spatial prepositions, connected split-screen compositions, functional UI-style prompt layouts, and text hierarchy for cleaner logo generation. A good example of how advanced prompt design can improve control without overcomplicating the wording.

Prompt Showcases

あみにん！ — 【PixAI / Tsubaki.2】ドストライクなｶﾜｲｲ娘の画像を作るためのプロンプト・設定・仕上げのコツまとめ (Tips for Creating Your Perfect Kawaii Character: Prompts, Settings & Finishing Touches)

This tutorial starts with a Grok-generated prompt and improves it through four practical stages: confirming direction, adding camera and pose detail, fixing breakdowns, and refining the background. Because every stage includes both the full prompt and the resulting image, it is one of the easiest workflows to reproduce directly.

@TOMBOAI7379 — Tsubaki.2 Neon Style: Prompt Showcase

A focused look at high-saturation neon aesthetics, using contrast-heavy color combinations and a dynamic “from above + heart finger + peek from inside” composition. A compact showcase of how Tsubaki.2 handles bold anime-style visual direction.

If you’d like to explore more prompting approaches, you can also refer to PixAI’s official Tsubaki.2 Prompt Guide. It works well alongside these community-created tutorials as an additional reference.

Multi-Character Generation & Consistency

Getting multiple characters to coexist in one image without visual blending is one of the hardest problems in anime AI image generation. These tutorials tackle that challenge directly.

所謂太郎 —PixAI最新モデルTsubaki2における、多人数キャラクターイラストの配置順・服装・ポーズの精密制御手法 ─ 実践的知見の共有 (Precise Control of Multi-Character Placement, Costume & Pose in Tsubaki.2)

The most ambitious project in the batch: ten named characters, one image, specific poses for each. What makes it especially valuable is its documentation of failure cases and root-cause analysis, along with direct comparisons against other anime image generation models. A standout contribution for anyone working on complex multi-character prompts.

Virtual-Rice-4039 — Good Prompt / Bad Prompt Species Mixing vs Species Separation Comma vs Period Prompting

A controlled experiment across three prompt versions, varying only the separator between character descriptions. Its conclusion — that periods create stronger semantic boundaries than commas when Prompt Helper is enabled — is one of the clearest examples in the challenge of small prompt structure changes producing meaningful gains in character separation.

@livybabie — Canon Anchors: Why Nami and Frieren Keep Failing

Introduces the “canon anchor” framework, arguing that character prompts often fail not because they are too short, but because they drift away from the character’s actual identity. The tutorial shows how the same logic can repair very different failure patterns, making it highly relevant for character consistency work

Text-in-Image Generation

Readable, correctly spelled text remains a specialized challenge in AI image generation. These tutorials offer some of the most practical guidance in the collection.

Virtual-Rice-4039 — How to Add Targeted Text in Tsubaki.2 Images — Why Quotation Marks Make It Stable

This tutorial focuses on one core method: wrapping the target word in double quotation marks, then describing its placement in natural language. It also documents what happens when the text is prompted without context and where the model’s limitations still appear. A strong reference for anyone trying to improve text rendering in generated images.

@yourself_life — Prompt Design: Text Generation + Multi-Character Poster

A three-step workflow for posters and title-based compositions: stabilize the text first, reinforce the logo structure if it breaks, then add character detail afterward. Because it solves a real spelling issue live, it is particularly helpful for users working on posters, title cards, and stylized promotional visuals.

Style Control in Tsubaki.2, LoRA & Aesthetic Directions

These tutorials explore how creators shape visual identity in Tsubaki.2 through Style Code, LoRA, and aesthetic planning.

perfect_ailove —Tsubaki.2 Style Code: Making Chibi Characters · Video

A practical walkthrough of using Style Code to lock in a chibi aesthetic without needing to engineer it through prompts. The fastest path from zero to a consistent character style.

@Yuki_Kyo_77 —關於：Tsubaki.2 與風格碼 (About: Tsubaki.2 and Style Codes)

A detailed reference on Style Code behavior, presented as a lighter and more model-compatible alternative to LoRA. It includes shareable style codes, a Claude-assisted prompt rewriting template, and multiple worked examples across genres. Especially useful for users looking to build reusable anime art workflows on PixAI.

Special Techniques

Costume & Apparel Generation

higurashihougi — Recovering Decorative Detail in Tsubaki 2: Why “More Maximal” Fails — and How Hierarchy Works Instead

This tutorial identifies an internal surface hierarchy in Tsubaki.2’s clothing generation, showing why adding more decorative keywords often fails on structural garment areas. Its proposed fix — redefining the role of the surface instead of intensifying descriptors — is a highly practical insight for ceremonial fashion, layered costumes, and complex apparel design.

@sui_AIcollege — Capturing the Cherry Blossom Spirit

A creative response to the model’s tendency to interpret petal-based clothing as woven fabric. By replacing fashion vocabulary with fluid-motion language and physically motivated structure, the tutorial shows how prompt framing can reshape material behavior in unexpectedly effective ways.

Manga & Comic Creation

@UNfukashigi — PixAI漫画生成プロンプトの作り方 (How to Write Manga Generation Prompts for PixAI)

A complete system for single-prompt manga page generation, with reusable structure, practical stability rules, and a free Google Sheet for assembling prompts. One of the strongest resources in the challenge for manga-oriented PixAI workflows.

斉川ニア — PixAi 簡単キャラクターデザインでアカウントを差別化しましょう (Differentiate Your Account with Easy Character Design in PixAI)

A full creative pipeline that moves from character sheet generation to annotation, story development, and manga output. More workflow-driven than theory-driven, and especially useful for creators thinking beyond single illustrations.

Tsubaki.2 Challenge Winners and Featured Tutorials

Taken together, these submissions did more than share isolated tips. They showed how creators are actually using Tsubaki.2 on PixAI across different goals, styles, and experience levels — from beginner-friendly anime prompting to advanced multi-character composition, text generation, and style control.

With that broader picture in mind, here are the winners from the Tsubaki.2 Tutorial Challenge.

A few notes:

Winning tutorials may be featured on the PixAI Official Blog with translations in English, Japanese, Korean, and Traditional Chinese. We will contact creators individually to discuss republication terms.
Video submissions were judged in a separate category.
Winners within each tier are listed in no particular order.

Congratulations to all of our winners!

First Prize — Articles (5 Selected)

First Prize — Video (1 Selected)

@teshio_k — PixAIの新フラグシップモデルTsubaki2の衝撃を徹底解説 (A Complete Breakdown of PixAI’s New Flagship Model Tsubaki.2)

Second Prize (10 Articles)

Second Prize — Video (1 Selected)

perfect_ailove —Tsubaki.2 Style Code: Making Chibi Characters

Participation Award

なるみなみ — PixAIチュートリアル Tsubaki.2 (Tsubaki.2 Tutorial for PixAI)
@teshio_k — Tsubaki.2 徹底活用ガイド：次世代アニメ画像生成の技術と実践 (Tsubaki.2 Complete Guide: Techniques & Practice)
@TOMBOAI7379 — Tsubaki.2 Neon Style: Prompt Showcase

What This Challenge Taught Us About Tsubaki.2

Across twenty submissions in three languages, several clear patterns emerged. Although each creator approached Tsubaki.2 differently, the strongest tutorials often arrived at similar conclusions about what makes prompts work — and where this model rewards precision most.

1. Prompt structure matters more than keyword density

One of the clearest patterns across the challenge was that stronger results usually came from better prompt structure, not from adding more descriptive words. Creators approached this in different ways — through step-by-step prompt frameworks, semantic separation tests, character anchoring, or even surface-role redefinition in clothing design — but the underlying idea was often the same: Tsubaki.2 responds better when the prompt defines relationships clearly.

In practice, that means composition, subject hierarchy, spatial separation, and functional roles often matter more than piling on extra adjectives. For creators using Tsubaki.2 on PixAI, this may be one of the most important prompt engineering takeaways from the challenge.

2. Tsubaki.2 rewards specificity

Another recurring theme was that Tsubaki.2 tends to preserve exactly what the prompt specifies — but does not reliably invent missing visual intent on its own. Multiple tutorials showed the same pattern from different angles: if lighting is not described, the image often defaults to flatter illumination; if the camera is not defined, composition becomes more generic; if character identity is loosely framed, visual drift becomes more likely.

That level of precision is part of what makes Tsubaki.2 powerful for anime AI image generation, but it also raises the bar for prompting. The more intentional the creator is about camera, lighting, text placement, or character relationships, the more controllable the result becomes.

3. The best workflows build images step by step

If there was one practical habit shared by almost every strong workflow tutorial, it was iteration. Rather than trying to solve composition, pose, style, lighting, background, and text all at once, creators consistently got better results by locking in direction first and then refining one variable at a time.

That pattern showed up in beginner guides, prompt showcases, poster workflows, and character design tutorials alike. For PixAI users working with Tsubaki.2, the takeaway is simple: start with a stable base, confirm the image is moving in the right direction, and then add complexity gradually. In many cases, that produces better results than trying to write the perfect all-in-one prompt from the beginning.

Closing Thoughts

Taken together, these tutorials did more than explain individual techniques. They showed how creators are learning to work with Tsubaki.2 as a model that rewards structure, specificity, and deliberate iteration. More importantly, they turned the challenge into a practical community resource — one that can help both new and experienced PixAI users build stronger workflows for anime image generation.

Thank you to everyone who joined the challenge and contributed their ideas, methods, and discoveries. We’re excited to see how these shared workflows continue to inspire new creations across the PixAI community.

Start Creating

Related Chapters

Tsubaki.2 Is Here — Everything You Need to Know