Anima Text-to-Image Workflow

This ComfyUI workflow is designed for Anima text-to-image generation, anime-style character creation, and direct prompt-based illustration output. The main purpose of this workflow is to let creators generate a complete image from text using Anima Preview, with a clean and simple node structure that is easy to understand, modify, and deploy online.

Unlike the Anima image-to-image workflow, this graph does not require an input image. It starts from an empty latent canvas and generates the image entirely from the user’s text prompt. This makes it suitable for original character design, anime illustration, concept art, social media cover images, Civitai example images, Bilibili thumbnails, YouTube visual assets, and quick creative testing.

The workflow is built around Anima Preview, using anima-preview.safetensors as the main diffusion model. It also uses qwen_3_06b_base.safetensors as the text encoder and qwen_image_vae.safetensors as the VAE. The structure is simple and direct: load the model, load the text encoder, encode the positive prompt, encode the negative prompt, create an empty latent image, run KSampler, decode the result, preview it, and save the final image.

The generation canvas is created with EmptyLatentImage. In the included setup, the latent size is 1024 x 1024 with batch size 1. This is a practical square format for anime character images, model previews, Civitai examples, and social platform assets. Users can adjust the width and height if they need portrait, landscape, or cover-style compositions.

The positive prompt is written directly in the CLIPTextEncode node. The included example prompt describes a high-quality anime-style solo female character with short black bob hair, amber eyes, a beauty mark under the left eye, silver cross earring, white military-style jacket with gold trim, black pleated skirt, black gloves, and black thigh-highs. This shows the workflow’s intended use: detailed character prompting with clear appearance, clothing, and visual direction.

The negative prompt is also included in the workflow. It suppresses common image-generation issues such as low quality, worst quality, blur, bad anatomy, bad hands, extra fingers, fused fingers, deformed face, text, watermark, logo, and JPEG artifacts. This is especially important for anime-style generation, where hands, fingers, facial details, and unwanted text artifacts can easily reduce the final quality.

The main sampling stage uses KSampler. In the uploaded setup, the workflow uses around 30 steps, CFG 4, er_sde sampler, simple scheduler, and full denoise. Since this is a pure text-to-image workflow, the denoise value is set to 1. This means the image is generated from the empty latent rather than being constrained by a source image. The seed is fixed in the example, which makes the output reproducible. Users can switch seed behavior to randomize if they want more variation.

After sampling, VAEDecode converts the generated latent into a visible image. The result is then passed into PreviewImage for quick checking and SaveImage for final export. This makes the workflow suitable for rapid prompt testing. Users can change the prompt, seed, resolution, sampler, or negative prompt and immediately compare different outputs.

This workflow is useful because it provides the simplest starting point for Anima Preview. If users want to test whether Anima fits their style needs, this text-to-image graph is the fastest route. It avoids extra complexity such as image input, auto-captioning, tiled upscale, inpainting, or reference conditioning. The workflow is clean enough for beginners, but still useful for advanced users who want a lightweight base graph to expand.

Creators can use this workflow to generate anime characters, stylized portraits, original character concepts, fashion designs, fantasy characters, game character concepts, visual novel characters, cover images, and social media assets. It is also a good base workflow for testing prompt language, model behavior, seed variation, composition control, and negative prompt strength.

Main features:

- Anima text-to-image workflow

- Uses anima-preview.safetensors

- Qwen 3 0.6B text encoder support

- Qwen image VAE support

- 1024 x 1024 empty latent generation

- Direct positive prompt input

- Negative prompt for anime artifact suppression

- KSampler generation route

- er_sde sampler with simple scheduler

- Full denoise text-to-image generation

- PreviewImage for quick checking

- SaveImage for final export

- Suitable for anime character creation and illustration testing

Recommended use cases:

Anima text-to-image generation, anime character design, original character creation, concept art testing, prompt research, Civitai example image generation, RunningHub workflow publishing, Bilibili cover creation, YouTube thumbnail assets, visual novel character design, game character concept art, social media artwork, and fast anime illustration prototyping.

Suggested workflow:

Start by editing the positive prompt. For character generation, describe the subject first, then the pose, clothing, hairstyle, facial features, mood, lighting, background, and rendering style. A clear prompt structure usually gives better results than a random list of tags.

For example, describe the character identity first: “1girl, solo, full body, dynamic pose.” Then add visual identity: “short black bob hair, amber eyes, beauty mark under left eye.” Then add clothing and accessories: “white military-style jacket with gold trim, black pleated skirt, black gloves, black thigh-highs.” Finally, add quality and style terms such as “masterpiece, best quality, highres, clean anime illustration.”

Use the negative prompt to control common problems. Keep terms such as low quality, blurry, bad anatomy, bad hands, extra fingers, fused fingers, deformed face, text, watermark, logo, and JPEG artifacts. If the model repeatedly creates a specific problem, add targeted negative terms.

Use seed control for repeatability. If you find a good character design, keep the seed fixed and adjust the prompt gradually. If the composition is not good, randomize the seed and test several versions. For character design, seed testing is often faster than rewriting the entire prompt.

Adjust resolution based on the target output. The included 1024 x 1024 setting is good for square previews and general testing. For vertical covers, change the canvas to a portrait ratio. For YouTube or Bilibili thumbnails, use a landscape ratio. Keep model-friendly dimensions and avoid extreme aspect ratios during early testing.

Use CFG carefully. The included CFG value around 4 is a balanced starting point. If the image does not follow the prompt well, increase CFG slightly. If the image becomes too rigid, noisy, or overcooked, reduce CFG. For Anima-style generation, moderate CFG usually gives cleaner results than extreme guidance.

Use 30 steps as a stable starting point. More steps may improve detail slightly, but also increase generation time. Fewer steps can be useful for fast testing. Once the prompt direction is good, use the stable step setting for final output.

After generation, inspect the result carefully. Check the face, hands, fingers, clothing details, body proportion, background, and overall composition. For anime character outputs, hands and small accessories often need special attention. If the result is close but not final, adjust the prompt and seed rather than changing every setting at once.

This workflow is designed as a clean Anima text-to-image starting point for ComfyUI users. It provides a direct route from prompt to final image, making it ideal for creators who want to test Anima Preview quickly, generate anime characters, build Civitai examples, or create reusable visual assets for content production.

🎥 YouTube Video Tutorial

Want to know what this workflow actually does and how to start fast?

This video explains what the tool is, how to launch the workflow instantly, and shares my core design logic — no local setup, no complicated environment.

Everything starts directly on RunningHub, so you can experience it in action first.

👉 YouTube Tutorial: https://youtu.be/J2A8JWDCUhk

Before you begin, I recommend watching the video thoroughly — getting the full context helps you understand the tool faster and avoid common detours.

⚙️ RunningHub Workflow

Try the workflow online right now — no installation required.

👉 Workflow: https://www.runninghub.ai/post/2021929707570208769/?inviteCode=rh-v1111

If the results meet your expectations, you can later deploy it locally for customization.

🎁 Fan Benefits: Register to get 1000 points + daily login 100 points — enjoy 4090 performance and 48 GB super power!

📺 Bilibili Updates (Mainland China & Asia-Pacific)

If you’re in the Asia-Pacific region, you can watch the video below to see the workflow demonstration and creative breakdown.

📺 Bilibili Video: https://www.bilibili.com/video/BV1FscqzREni/

☕ Support Me on Ko-fi

If you find my content helpful and want to support future creations, you can buy me a coffee ☕.

Every bit of support helps me keep creating — just like a spark that can ignite a blazing flame.

👉 Ko-fi: https://ko-fi.com/aiksk

💼 Business Contact

For collaboration or inquiries, please contact aiksk95 on WeChat.

🎥 YouTube 视频教程

想了解这个工作流到底是怎样的工具，以及如何快速启动？

视频主要介绍 工具定位、快速启动方法 和 我的构筑思路。

我们会直接在 RunningHub 上进行演示，让你第一时间看到实际效果。

👉 YouTube 教程： https://youtu.be/J2A8JWDCUhk

开始前建议尽量完整地观看视频 —— 把握整体思路会更快上手，也能少走常见弯路。

⚙️ 在线体验工作流

现在就可以在线体验，无需安装。

👉 工作流： https://www.runninghub.ai/post/2021929707570208769/?inviteCode=rh-v1111

打开上方链接即可直接运行该工作流，实时查看生成效果。

如果觉得效果理想，你也可以在本地进行自定义部署。

🎁 粉丝福利： 注册即送 1000 积分，每日登录 100 积分，畅玩 4090 体验 48 G 超级性能！

📺 Bilibili 更新（中国大陆及南亚太地区）

如果你在中国大陆或南亚太地区，可以通过下方视频查看该工作流的实测效果与构思讲解。

📺 B站视频： https://www.bilibili.com/video/BV1FscqzREni/

我会在 夸克网盘 持续更新模型资源：

👉 https://pan.quark.cn/s/20c6f6f8d87b

这些资源主要面向本地用户，方便进行创作与学习。