CivArchive
    Tifa WAN2.2 T2V - v1.0
    NSFW

    Brief descriptions don't seem to work for this LORA.

    Place the prompt into Gemini 2.5 Pro or another language model:

    “遵循以下公式要求,参考示例,补全输入的提示词,只需要用英文输出提示词。

    提示词公式:主体(主体描述)+ 场景(场景描述)+ 运动(运动描述)+ 美学控制

    1.主体描述:主体描述是对主体外观特征细节的描述,可通过形容词或短句列举,例如“一位身着少数民族服饰的黑发苗族少女”、“一位来自异世界的飞天仙子,身着破旧却华丽的服饰,背后展开一对由废墟碎片构成的奇异翅膀”。

    2.场景描述:场景描述是对主体所处环境特征细节的描述,可通过形容词或短句列举。

    3.运动描述:运动描述是对运动特征细节的描述,包含运动的幅度、速率和运动作用的效果,例如“猛烈地摇摆”、“缓慢地移动”、“打碎了玻璃”。

    4.美学控制:包含光源、光线环境、景别、视角、镜头、运镜等,常见镜头语言详见下方提示词词典。

    5.把tifa描述为“女人tifa”。

    提示词示例:

    1.边缘光,低对比度,中近景,日光,左侧重构图,干净的单人镜头,暖色调,柔光,晴天光,侧光,白天,一个年轻的女孩坐在高草丛生的田野中,两条毛发蓬松的小毛驴站在她身后。女孩大约十一二岁,穿着简单的碎花裙子,头发扎成两条麻花辫,脸上带着纯真的笑容。她双腿交叉坐下,双手轻轻抚弄身旁的野花。小毛驴体型健壮,耳朵竖起,好奇地望着镜头方向。阳光洒在田野上,营造出温暖自然的画面感。

    类型:

    1.光源类型:日光、人工光、月光、实用光、火光、荧光、阴天光、混合光、晴天光

    2.光线类型:柔光、硬光、顶光、侧光、背光、底光、边缘光、剪影、低对比度、高对比度

    3.时间段:白天、夜晚、黄昏、日落、黎明、日出

    4.景别类型:特写、近景、中景、中近景、中全景、全景、广角

    5.构图类型:中心构图、平衡构图、右/左侧重构图、对称构图、短边构图

    6.镜头焦段类型:中焦距、广角、长焦、望远、超广角-鱼眼

    7.机位角度:过肩角度、高角度、低角度、倾斜角度、航拍、俯视角度

    8.镜头类型:干净的单人镜头、双人镜头、三人镜头、群像镜头、定场镜头

    9.色调类型:暖色调、冷色调、高饱和度、低饱和度

    10.基础运镜类型:镜头向右移动、镜头向左移动、镜头上摇

    11.高级运镜类型:手持镜头”.

    This prompt is copied from the official website and has been briefly summarized.

    For example:

    input: Tifa is drinking at the bar counter.

    Output: Edge light, low contrast, medium close-up shot, artificial light, right-heavy composition, clean single shot, warm color tone, soft light, night, low angle, a woman Tifa with long dark hair, wearing her iconic white tank top and black leather skirt, is drinking at a polished dark wood bar counter. The counter reflects the dim neon lights from the bar. She slowly raises a glass of amber liquor to her lips, her expression is contemplative.

    I have only tested it with low-noise lightx2v lora.

    recommended not to use high-noise lightx2v LoRa.

    euler+simple

    81 frames:

    high-noise steps:5

    high-noise model cfg:4

    low-noise steps:3

    low-noise model cfg:3

    shift:12

    41 frames :

    euler+simple

    high-noise steps:3

    high-noise model cfg:4

    low-noise steps:3

    low-noise model cfg:3

    shift:12

    If the generated video quality is unsatisfactory, try adding a VideoLinearCFGGuidance node with a value of 1 before the low-noise ksampler input model.

    Description

    Just the character Tifa

    LORA
    Wan Video 2.2 T2V-A14B

    Details

    Downloads
    24
    Platform
    SeaArt
    Platform Status
    Available
    Created
    9/16/2025
    Updated
    9/16/2025
    Deleted
    -

    Files

    Available On (1 platform)

    Same model published on other platforms. May have additional downloads or version variants.