热门工作流

上传一张含主角的人物原图和一张目标背景图，自动抠出人物并自然融入新场景。严格保留五官、发型、服装、配饰、姿态和身体比例，不做改脸或美化；匹配背景的构图、透视、光照方向与色温，边缘自然、无白边、无重影。一次输出一张写实合成图，无需填写提示词。

上传商品主图、模特图、姿势图与环境图，一次生成模特穿着该商品的成片，并附一张袖口/裤口的局部工艺特写；同一行还会给出姿势白模、空场景与模特身份参考，便于人工复核。

上传一部网文全文（.txt/.md），输出一份 8 节的《结构拆解报告》：结构基线表、逐章转折点密度图、功能性角色拓扑、爽点在全文的百分比位置、开篇钩子拆解、三段式节奏诊断、可跨题材复用的手法清单、本次报告的局限。所有字数、章节数、对话占比等数字由程序确定性统计，模型只做结构判断与成文——不复制原文、不代写、不做文学评价。单次分析上限 30 章 / 6 万字，超出部分按章节边界截断，并在报告首行明确告知。

上传同一成年模特的1至4张原始参考图，填写全身、头像或其他视图要求，生成一张模特设定候选。请使用获授权素材，检查脸部身份、身体比例和手脚后再保存复用。已完成一张正面全身参考的真实生成与人工初检；身份与其他视角仍需逐图复核。 创作者附加费用为0，模型运行费用按平台实际计费；运行前查看费用提示。

每行输入一款商品原图、模特参考、场景参考、姿势参考、可选商品细节图、拍摄要求及本款商品结构约束，生成一张穿着候选。商品原图与细节图优先定义服装，姿势图仅提供动作。结构约束请明确袖口、裤口、领口和缝线等可见特征，逐款填写，避免套用其他款式。已完成单款六图人工确认；另一款四图待审、两行被模型输出审核拒绝。需逐图复核商品细节，不保证批量完全一致。 创作者附加费用为0，模型运行费用按平台实际计费；运行前查看费用提示。

上传环境图，可选上传道具图，并填写场景、机位和留白要求，生成空景候选，供后续服装商品图复用。请检查道具数量、透视、光源方向与人物站位空间。已完成单样例运行，复杂空间一致性需逐图复核。 创作者附加费用为0，模型运行费用按平台实际计费；运行前查看费用提示。

上传一张姿势照片，生成去除人物身份、服装纹理和环境的中性白模姿态参考。保留可见肢体方向、接触关系和原图景别；不补全被裁切身体。输出为二维近似参考，需人工检查左右、手指和支撑物后再复用。已完成站姿、侧身及袖口展示的小样本运行。 创作者附加费用为0，模型运行费用按平台实际计费；运行前查看费用提示。

上传一件服装图和一句场景，出持杆、行走、近景三张。

上传新商品图和 4 张成片，按原版式换成新商品。

Generate a short material breakdown ad from a one-sentence material selling point.

Generate a short minimalist product ad from a one-sentence product description.

Describe your brand and product in one sentence to generate a short brand promo video.

Upload a product image and a one-sentence idea to create a scene image and a main image with a price tag.

Upload photos and choose from 8 styles for each row: travel sketch, photo doodle, palette-knife painting, 3D paper sculpture, rough relief, real-world fairytale, and two anime styles. Generate one image per row and compare multiple styles for the same photo. Creator fees follow the marketplace quote; model image-generation fees are charged separately.

Paste your text and choose the art style, layout, colors, and narrative approach to create five 3:4 portrait cards for Xiaohongshu: a cover, three content pages, and a closing page. Uses a fixed model with no creator fee. Inspired by baoyu-xhs-images from JimLiu/baoyu-skills (MIT).

Paste your text and upload reference images to automatically create five 3:4 portrait cards for Xiaohongshu: a cover, three content pages, and a closing page. Uses a fixed model with no creator fee. Inspired by baoyu-xhs-images from JimLiu/baoyu-skills (MIT).

Paste your text to automatically create five 3:4 portrait cards for Xiaohongshu: a cover, three content pages, and a closing page. Uses a fixed model with no creator fee. Inspired by baoyu-xhs-images from JimLiu/baoyu-skills (MIT).

AI Short Drama Studio; ¥2 per run (up to 30 seconds); v0.5.0 includes qc/skill_check/build_release.

Enter a video prompt, select a video model and a 16:9 or 9:16 aspect ratio for each row, and generate MP4 video assets.

Generate short, silent video clips suitable for product demonstrations based on text scripts.

Paste the text of your article, and AI will use 6 storytelling methods to generate 5 story-type reference titles for you: Hero’s Journey/Conflict-Suspense-Unexpected/About the Other Side/Use Pictures to Replace Data/Golden 7 Seconds Opening/Concrete Details. Each title clearly labels which method was used, turning the methodology from paper into a selectable title. Based on Claude Opus 4.8.

Paste a WeChat official account article and get title suggestions from 7 different perspectives at once: from emotional resonance, suspense attraction, digital usefulness, counter-intuitive controversy, questioning style to SEO keywords, each of the 6 different models will give you a best title, and then the editor will independently recommend one. Choosing a title is no longer based on feeling, but after seeing the complete perspective.

Deeply rewrite the AI text into the original style of real people, reduce the AI rate and pass AIGC detection. It has built-in 14 categories of AI word removal, sentence pattern demolding, six major platform writing styles, GEO optimization, and supports terminology locking and healthy track compliance. ¥0.99/time.