EasyHu21のインストール 及びggufモデルのHunyuanImage2.1 とHunyuanVideo1.5のワークフロー

はじめに

HunyuanImage-2.1 を簡単に試せる環境です。

インストール 起動

GitHub – Zuntan03/EasyHu21: Easy HunyuanImage-2.1

EasyHu21Installer.bat を右クリックして 名前をつけてリンク先を保存 。

インストール先の 空フォルダ を 作成。D:/EasyHu21/ などに用意して、EasyHu21Installer.bat を置いて実行します。

しばらく待つとインストールが完了します

ComfyUi.bat で起動します。

Hu21/00-TextToImage ワークフローを開いて 実行する 。

画像が作成されたら OK です

アウトオブメモリーのエラーが出た時はComfyUI-MultiGPU を使っているので調整します

ComfyUiの簡単な使い方については次のページが参考になります

HunyuanImage-2.1 別なシンプルなワークフローを使う

参考

ComfyUIでFP8/GGUF版のHunyuanImage-2.1 Distilledを試す(推奨VRAM 12~16GB)|まゆひらa

流れ

LoaderGGUF (ID:129)  ← モデルの読み込み
   ↓
KSampler (ID:16)  ← 潜在画像生成
   ↓
VAEDecode (ID:21)  ← 画像に変換
   ↓
SaveImage (ID:66)  ← 保存

モデル

ComfyUI/
├── models/
│   ├── vae/
│   │   └── HunyuanImageVae-fp16.safetensors         ← VaeGGUF ノード(ID:128)で使用
│   ├── loras/
│   │   └── (このワークフローでは未使用)
│   ├── unet/
│   │   └── HunyuanImage21_distilled-Cc-Q4_K_M.gguf  ← LoaderGGUF ノード(ID:129)で使用
│   ├── text_encoders/
│   │   └── Qwen25VL-7B-Instruct-UD-Us-Q4_K_XL.gguf  ← DualClipLoaderGGUF ノード(ID:130)の clip_name1
│   │   └── ByT5Small_GlyphSdxl2-fp32-Cc-F16.gguf     ← DualClipLoaderGGUF ノード(ID:130)の clip_name2

プロンプト

チャッピーに作ってもらいました



Bathed in the golden light of early morning, the young girl sits quietly by a sunlit window, nestled in a cozy room adorned with soft curtains and shelves lined with well-loved books and tiny potted plants. The sunlight streams through the glass, casting warm, dappled patterns across her delicate features and the polished wooden floor.

Her long, dark brown hair tumbles in gentle waves over her shoulders, catching the morning glow. A shy smile plays on her lips as she rests her cheek against her hand. She’s dressed in a flowing, cream-colored dress with lace trim and a ribbon tied at the waist, the fabric pooling softly around her as she sits. The gentle rustle of the dress as she shifts in her seat adds to the tranquil hush of the room.

In front of her, a steaming cup of coffee rests on a small round table, sending up delicate wisps of warmth that curl into the cool morning air. Outside the window, the sleepy mountain village begins to stir?stone rooftops glinting in the light, a few birds flitting between trees, and the distant sound of a bell marking the start of a new day.

Above her head, floating gently in the golden light, is a soft speech bubble. In delicate, handwritten script, containing the Japanese text "おはよう" with a line break as specified.

 a simple, heartfelt greeting, as if she’s inviting you to sit beside her, share the warmth, and begin the day together. ???

ワークフロー

HunyuanVideo1.5

モデル

📂 ComfyUI/
├── 📂 models/
│   ├── 📂 text_encoders/
│   │      ├── Qwen2.5-VL-7B-Instruct-Q4_K_M.gguf
│   │      ├── Qwen2.5-VL-7B-Instruct-mmproj-BF16.gguf
│   │      └── ByT5Small_GlyphSdxl2-fp32-Cc-F16.gguf
│   ├── 📂 clip_vision/
│   │      ├── sigclip_vision_patch14_384.safetensors
│   ├── 📂 diffusion_models/
│   │      ├── hunyuanvideo1.5_480p_i2v_cfg_distilled-Q4_K_M.gguf
│   │      └── hunyuanvideo1.5_480p_t2v_cfg_distilled-Q4_K_M.gguf
│   └── 📂 vae/
│          └── hunyuanvideo15_vae_fp16.safetensors

https://huggingface.co/jayn7/HunyuanVideo-1.5_I2V_480p-GGUF

https://huggingface.co/jayn7/HunyuanVideo-1.5_T2V_480p-GGUF

https://huggingface.co/Comfy-Org/HunyuanVideo_1.5_repackaged/tree/main/split_files

プロンプト

Text-to-Video

  • Basic Usage: Subject + Motion + Scene
  • Advanced Usage: Freely add more control tags, e.g., Subject + Motion + Scene + Style + Camera Movement + Lighting

https://github.com/Tencent-Hunyuan/HunyuanVideo-1.5/blob/main/assets/HunyuanVideo_1_5_Prompt_Handbook_EN.md

ワークフロー作成中

ComfyUI

Posted by eightban