Shao Meng: Most domestic Video Agents have been reduced to token relay stations

Brothers, Recently the "teleport sliding mirror" videos on Douyin have been extremely popular 🔥 I tried to replicate one using a white-model video. This time I chose ZenMux @ZenMuxAI to get it all done in one stop. I've introduced ZenMux many times before—it's an enterprise-level model aggregation service platform, with one account and one set of APIs, connecting text, image, and video models from multiple vendors. It only takes 3 steps to complete the replication of this viral video: - Use GPT Image 2.5 to generate two three-view images of Malfoy, before and after - Convert the reference video into a white-model video - Use the reference images and the white-model video, plus the Seedance 2.5 prompt, and you're done I used only one API for the entire process and got everything done, without switching back and forth between different platforms—very convenient. --- Coinciding with Halloween 🎃, ZenMux is doing something interesting this time—use AI to get work done and open a model blind box along the way. With cash pay-as-you-go (PAYG) spending of $1 or more, you can go to the event page to unlock a blind box, then open it to view the reward. If you have code to modify, prompts to optimize, or images and videos to make, you can participate with real tasks. After opening a model reward, try the same task again and see what it can actually help you with. This time the prize pool has 16 models, with 3,000 rewards per day, including Claude, GPT, Gemini, DeepSeek, Kling, etc. There is about a 44% chance of opening Claude / GPT / Gemini, and a single reward has a maximum usage value of $500. The event runs from 22:00 on October 26 to 22:00 on October 31, refreshing every night at 22:00. Event entry: https://t.co/oXOL8pQXgy --- The video prompt is below 👇👇👇

View original

兄弟们， 最近抖音上的「瞬移滑镜」的视频非常火爆🔥 尝试用白膜视频复刻一下。 这次我选的是ZenMux @ZenMuxAI 一条龙搞定，ZenMux我之前介绍过多次了，这是一个企业级模型聚合服务平台，一个账户、一套 API，接入多家厂商的文本、图片和视频模型。 只需3步就可以完成这个爆款视频的复刻： - 用GPT Image 2.5生成Malfoy前后两张三视图 - 将参考视频转成白膜视频 - 使用参考图和白膜视频，加上Seedance 2.5提示词就可以了 我全程只用了一个API就全部搞定了，不用在不同的平台来回切换，非常方便。 --- 恰逢万圣节🎃，ZenMux这次有点意思，用 AI 干活，还能顺手开个模型盲盒 现金按量付费（PAYG）消费满 $1，就能去活动页解锁盲盒，再开盒查看奖励。 手头如果有代码要改、提示词要优化、图片和视频要做的朋友，可以拿真实任务参与。 开到模型奖励后，再拿同一个任务试试，看看它究竟能帮上什么忙。 这次奖池有 16 款模型，每天 3,000 份奖励，包含 Claude、GPT、Gemini、DeepSeek、Kling 等，约 44% 的概率开到 Claude / GPT / Gemini，单份奖励最高使用价值 $500。 活动时间是10 月 26 日 22:00 到 10 月 31 日 22:00，每晚 22:00 刷新。 活动入口：https://t.co/oXOL8pQXgy --- 视频提示词见下面👇👇👇