README.md

---
license: other
license_name: civitai-model-license
license_link: https://civitai.red/models/2764727/tripleview
pipeline_tag: image-to-image
tags:
  - lora
  - comfyui
  - flux.2
  - flux.2-klein-9b
  - krea-2
  - character-sheet
  - character-design
  - image-editing
  - multi-view
---

# CharacterSheet — TripleView & QuadView LoRAs

CharacterSheet is a collection of image-editing LoRAs that converts an existing character image into a clean, multi-view reference sheet. The models are designed to preserve the character's identity, clothing, proportions, and overall appearance across the generated views.

The collection was trained on approximately 300 character-sheet examples. Its main training focus was human characters, although anime, cartoon, animal, creature, and other stylized subjects can also work. Results outside the primary training distribution may be less consistent.

The generated sheets are useful for:

- Character design and turnarounds
- Front, side, and back reference views
- Character dataset preparation
- CharacterID workflows
- LTX identity-reference pipelines
- Further manual or generative refinement

> These are LoRAs, not standalone checkpoints. Use each file with its matching base model and workflow.

## Available models

| File | Base model | Output layout | Included workflow |
| --- | --- | --- | --- |
| `TripleView_klein9b_v1.safetensors` | FLUX.2 Klein 9B | Front, side, and back full-body views | `workflows/TripleView_klein9b_v1.json` |
| `QuadView_klein9b_v1.safetensors` | FLUX.2 Klein 9B | Face close-up plus front, side, and back full-body views | `workflows/QuadView_klein9b_v1.json` |
| `QuadView_krea2_v1.safetensors` | Krea 2 | Face close-up plus front, side, and back full-body views | `workflows/QuadView_krea2_v1.json` |

## Usage in ComfyUI

1. Download the LoRA that matches your base model.
2. Place it in `ComfyUI/models/loras/`.
3. Open the corresponding JSON file from the `workflows` folder in ComfyUI.
4. Load a clear reference image of the character.
5. In the LoRA Loader node, select the downloaded file. The filename stored in the workflow may differ from the filename in this repository.
6. Use the prompt for the selected layout, then queue the workflow.

### Trigger captions

**TripleView — FLUX.2 Klein 9B**

```text
Convert the character in the image to a Character Sheet showing front, side and back full body views
```

**QuadView — FLUX.2 Klein 9B**

```text
Convert the character in the image to a Character Sheet showing a face close-up, front, side and back full body views
```

**QuadView — Krea 2**

```text
Convert the character in the image to a Character Sheet showing a face close-up, front full body, side full body and back full body views
```

## Recommended starting settings

These values match the included workflows and are intended as practical starting points.

| Variant | Resolution | Steps | CFG | Sampler | Scheduler | LoRA strength |
| --- | ---: | ---: | ---: | --- | --- | ---: |
| TripleView — FLUX.2 Klein 9B | 1536×1024 | 8 | 1 | Euler | beta57 | 1.0 |
| QuadView — FLUX.2 Klein 9B | 1536×1024 | 8 | 1 | Euler | simple | 1.0 |
| QuadView — Krea 2 Turbo | 1536×1024 | 10 | 1 | Euler | simple | 1.0 |

For Krea 2 Turbo, the included workflow notes that 8 steps can favor stronger composition or instruction adherence, while 12 steps can improve facial detail. When using Krea 2 Raw instead of Turbo, start around 40 steps with CFG 3–4.

## Examples

The examples below show the reference input and generated QuadView result using the Krea 2 variant.

<p align="center">
  <img src="samples/krea2/ComfyUI_temp_mlgmd_00115_.png" alt="Krea 2 QuadView sample 1" width="100%">
</p>

<p align="center">
  <img src="samples/krea2/ComfyUI_temp_mlgmd_00116_.png" alt="Krea 2 QuadView sample 2" width="100%">
</p>

<p align="center">
  <img src="samples/krea2/ComfyUI_temp_mlgmd_00117_.png" alt="Krea 2 QuadView sample 3" width="100%">
</p>

<p align="center">
  <img src="samples/krea2/ComfyUI_temp_mlgmd_00120_.png" alt="Krea 2 QuadView sample 4" width="100%">
</p>

<p align="center">
  <img src="samples/krea2/ComfyUI_temp_mlgmd_00123_.png" alt="Krea 2 QuadView sample 5" width="100%">
</p>

<p align="center">
  <img src="samples/krea2/ComfyUI_temp_mlgmd_00128_.png" alt="Krea 2 QuadView sample 6" width="100%">
</p>

<p align="center">
  <img src="samples/krea2/ComfyUI_temp_mlgmd_00131_ (1).png" alt="Krea 2 QuadView sample 7" width="100%">
</p>

<p align="center">
  <img src="samples/krea2/ComfyUI_temp_mlgmd_00132_.png" alt="Krea 2 QuadView sample 8" width="100%">
</p>

<p align="center">
  <img src="samples/krea2/ComfyUI_temp_mlgmd_00136_.png" alt="Krea 2 QuadView sample 9" width="100%">
</p>

<p align="center">
  <img src="samples/krea2/krea2_identity_edit_00293_.png" alt="Krea 2 QuadView sample 10" width="100%">
</p>

## Limitations

- The model can introduce differences in anatomy, facial features, clothing, accessories, or proportions between views.
- Side and back views may require additional generations or manual cleanup.
- Human characters were the primary training focus; stylized characters and non-human subjects may be less predictable.
- Detailed accessories, asymmetric designs, hands, and complex garments can be difficult to preserve exactly.
- A clear, well-framed input image generally produces a more consistent sheet.

## License and source

Usage permissions follow the [original Civitai model listing](https://civitai.red/models/2764727/tripleview). Review that page before redistribution or commercial use.

If this project is useful to you, you can [support continued model development](https://buymeacoffee.com/nrdx).