Version 3 has less impact on the default style of official ZIT models and does not override the characters built into the model. It is also compatible with character LoRAs trained based on ZIB.

In addition, it provides more flexible adjustments for a wide range of age variations.

--

2026/01/30:BASE V3

V3版本基於Z Image Base訓練，可在Z Image Base和Z Image Turbo上使用。

在Z Image Base上使用時建議下載Z-Image-Fun-Lora-Distill-4-Steps-2602-ComfyUI lora，4步即可生成圖片。

解析度為1280p，使用H200 SXM訓練。

V3版本沒有訓練人臉，不會影響Z Image中的內建人物，並且對於Z Image的原始風格影響較小。

建議將A和B版本都下載來使用，這樣組合起來會更靈活。

V3分為A和B版本，A版本訓練步驟較低，B版本訓練步驟較高。

在Z Image Turbo上使用時:

當A和B一起使用時權重都設定為”1.25~1.75”，當A或B單獨使用時，權重設定為”2~3”。

V3版本lora在原版的Z Image Turbo上使用時，可與人物lora一起使用。

(注意:我只測試我自己練的人物lora.)

在Z Image base上使用時:

當A和B一起使用時權重都設定為”0.6”，當A或B單獨使用時，權重設定為”0.8~1”。

The V3 version is trained based on Z Image Base and can be used with both Z Image Base and Z Image Turbo.

When using it with Z Image Base, it is recommended to download the Z-Image-Fun-Lora-Distill-4-Steps-2602-ComfyUI LoRA, which can generate images in just 4 steps.
The resolution is 1280p, trained using H200 SXM.

The V3 version does not include face training, so it will not affect the built-in characters in Z Image and has minimal impact on Z Image’s original style.

It is recommended to download and use both the A and B versions for more flexible combinations.
V3 is divided into A and B versions: the A version is trained with fewer steps, while the B version is trained with more steps.

When using with Z Image Turbo:

· If using A and B together, set both weights to “1.25–1.75”.

· If using A or B individually, set the weight to “2–3”.

· When the V3 LoRA is used with the original Z Image Turbo, it can be combined with character LoRAs.
(Note: I have only tested this with my own trained character LoRAs.)

When using with Z Image Base:

· If using A and B together, set both weights to “0.6”.

· If using A or B individually, set the weight to “0.8–1”.

--

2025/12/05: V2

A lora或B lora可單獨使用，A訓練步驟較低，B訓練步驟較高，當A和B一起使用時只需LORA權重0.4就可生效。

但是要組合和其它LORA時，則必須A或B單獨使用。

此LORA會影響人臉，如果要還原內建人物的臉則必須進行面部修復。

或是使用A和B一起使用並且權重0.4即可降低對內建人物臉部的影響。

使用H200 SXM訓練。

--

Lora A or Lora B can be used individually. Lora A has a lower training step count, while Lora B has a higher training step count. When A and B are used together, a Lora weight of 0.4 is sufficient for effectiveness.

However, to combine them with other Lora types, either A or B must be used individually.

This Lora type affects the face; facial retouching is necessary to recreate the built-in character's face.

Alternatively, using A and B together with a weight of 0.4 can reduce the impact on the built-in character's face.

Training is performed using H200 SXM.

--

觸發詞(Trigger words):

pornmaster flat breasts, a woman with flat chest,

--

我的其他檢查站和Lora。

My other checkpoints and Lora.

--

客製化lora:我的Telegram ID:@iamddtla0620

我的微信ID:iamddtla

我的line ID:iamddtla

時間寶貴，請直接說明來意，例如:訂製lora.

Customized lora: My Telegram ID: @iamddtla0620

My WeChat ID: iamddtla

My Line ID: iamddtla

Time is precious, please state your intention directly, for example: customized LORA.

Search AI Models & Addons

very few results