Anima Preview2 Text-to-Image Workflow

This workflow is designed for Anima Preview2 text-to-image generation, with a simple but practical two-stage structure for creating polished anime-style images from prompts. Its main purpose is to let creators generate a clean base image with Anima Preview2, then refine and enlarge the result through a second latent pass for better detail, stronger visual finish, and more useful showcase output.

The workflow uses anima-preview2.safetensors as the main generation model, qwen_3_06b_base.safetensors as the text encoder, and qwen_image_vae.safetensors as the VAE. It starts from an EmptyLatentImage at 1024 x 1024, which makes it suitable for square-format anime illustrations, Civitai preview images, social media posts, character concept tests, and RunningHub online demos. Since the workflow is fully text-to-image, no input image is required. The user only needs to edit the positive and negative prompts, then run the graph.

The positive prompt in the workflow is built for cinematic anime illustration. It describes a 2025-style high-quality anime movie still with a solo original character, long brown hair, amber eyes, a vintage coat, a pocket watch, a train interior, cherry blossoms, embers, night reflections, elegant soft light, emotional atmosphere, and a detailed background. This kind of prompt is useful for testing whether Anima Preview2 can handle character beauty, atmosphere, lighting, background storytelling, and cinematic framing at the same time.

The negative prompt is also practical. It suppresses common generation issues such as low quality, worst quality, blur, bad anatomy, bad hands, extra fingers, fused fingers, deformed faces, text, watermark, logo, and JPEG artifacts. This helps keep the output cleaner, especially when the workflow is used for repeated prompt testing or public preview examples.

The first KSampler stage uses a 30-step generation pass with DPM++ 2M SDE GPU sampling, SGM Uniform scheduling, CFG around 3, and full denoise. This stage creates the main image from the prompt and defines the composition, character identity, lighting, color mood, and background structure. The result is decoded through the Qwen image VAE and previewed for inspection.

After the first image is created, the workflow adds a second refinement stage. The generated image is scaled through image_scale_pixel_v2 at 1.6x, encoded back into latent space with VAEEncode, and processed by a second KSampler pass. This second pass uses 20 steps, CFG around 2.5, and denoise around 0.65. The purpose is not to completely redraw the image, but to enhance it: cleaner details, richer texture, better line quality, and a more polished final result.

This makes the workflow more useful than a single-pass text-to-image setup. It gives creators a direct route from prompt to refined output without needing a separate upscaler workflow. It is suitable for anime character generation, cinematic illustration, prompt testing, Civitai showcase images, thumbnail concepts, and high-quality RunningHub workflow publishing.

In short, this is a compact Anima Preview2 text-to-image workflow with built-in second-pass refinement. If you want to see how the prompt, two-stage sampling, image scaling, and final detail pass are connected, watch the full tutorial from the YouTube link above.

⚙️ Try the Workflow Online

👉 Workflow: https://www.runninghub.ai/post/2033542967683715074?inviteCode=rh-v1111

Open the link above to run the workflow directly online and view the generation results in real time.

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

🎁 Fan Benefits: Register now to get 1000 points, plus 100 daily login points — enjoy 4090-level performance and 48 GB of powerful compute!

📺 Bilibili Updates (Mainland China & Asia-Pacific)

If you are in Mainland China or the Asia-Pacific region, you can watch the video below for workflow demos and a detailed creative breakdown.

📺 Bilibili Video: https://www.bilibili.com/video/BV1Q1w1zKEwk/

I will continue updating model resources on Quark Drive:

👉 https://pan.quark.cn/s/20c6f6f8d87b

These resources are mainly prepared for local users, making creation and learning more convenient.

⚙️ 在线体验工作流

👉 工作流： https://www.runninghub.ai/post/2033542967683715074?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/BV1Q1w1zKEwk/

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。