AI & Automation Enthusiast🤖 Aspiring UI/UX Designer 🎨 Learning Figma & sharing my progress. Building in public, one design at a time. 🚀
Joined September 2024
- Tweets1.6K
- Following1K
- Followers983
- Likes455
I’m conducting a quick UI/UX survey and would love to hear your thoughts. Your feedback will help me understand different user perspectives. It only takes a minute.
I’d appreciate your participation and a repost. 🙏
#UIUX #UXResearch #ProductDesign #DesignTwitter #UXDesign
Clue less retweeted
Of all the rewards in the ZenMux's Halloween event, $50 in PAYG credits feels like the most flexible one, because it isn't tied to a single model.
A pile of tokens for one model only helps if that's the one you need, whereas credits could go on whatever the week throws up.
I'd be interested to know what your first spend will be, and whether the credits work across text, image and video, or only some of them. The expiry date on yours is the other thing I'd check straight away, since that decides how fast you need a plan.
I joined ZenMux’s Halloween event, and my reward was $50 in PAYG credits. 🎃
That’s the reward shown in my screenshot. Results vary, and these are usage credits that cannot be exchanged for cash.
The event works like this: spend at least $1 in eligible cash pay-as-you-go usage during an event day on an individual ZenMux account to unlock a mystery-box opportunity.
If you already have an AI task to work on, your eligible usage can also count toward the Halloween promotion.
And if you’ve wanted to explore different AI models without taking out a separate subscription for each one, ZenMux offers access to multiple providers through one account, with pay-as-you-go usage available. You can choose among supported text, image, and video models to suit the task, then switch when you want to try something different.
A few details before joining:
• One opportunity per individual account per event day; unused opportunities don’t carry over.
• Subscription usage, promotional credits, gifted balances, and usage paid for with rewards don’t count toward eligibility.
• Each event day resets at 14:00 UTC. The event ends October 31 at 14:00 UTC.
• Check the expiry date and usage conditions shown on your own reward.
For the event details and rules, visit the official campaign page:
[zenmux.ai/halloween](zenmux.ai/halloween)
If you’d like to join through my invitation:
[zenmux.ai/invite/UNMMLS](zenmux.ai/invite/UNMMLS)
Clue less retweeted
The "ten shots, ten faces" problem is the one I'd worry about most too. Step 3 looks like it does the heavy lifting, since a character that stays put across 500 episodes is mostly a reference sheet problem, not a video model problem.
I also like that step 5 says to redo only the broken shot, not the whole sequence. That's probably where a two-person team saves most of its time.
卧槽 Σ(°ロ°)！500 多集？全网 41 亿播放？就两个人？这得赚多少钱？！
四川一对夫妻，在出租房卧室里手搓红果、绿泡泡 AI 短剧。
网传红果短剧年化收入逼近 400 亿元，不过这个数字已被官方辟谣。
营收传闻不靠谱，但有个问题值得研究：AI 能不能让两台电脑、一个小团队，也跑出一条内容生产线？
如果新闻里的数据属实，这个案例真正值得研究的，不只是 AI 能生成什么，而是一个小团队怎么把内容生产流程跑起来。
做 AI 视频最头疼的，可能不是生成不出来，而是同一个主角，生成十个镜头就换了十张脸。
拆一条 AI 短剧工作流，大致分五步：
1️⃣ 剧本拆分： 用 GPT、Claude 等文本模型拆剧情、写分镜，明确镜头动作、情绪、台词和景别。
2️⃣ 角色定妆： 用 GPT Image、FLUX 等生图模型生成角色设定图，固定脸型、发型和服装。
3️⃣ 多角度参考： 补充正面、侧面、全身等参考图，也可以用 ComfyUI 搭建可复用的角色出图流程，减少人物漂移。
4️⃣ 视频生成： 把角色参考图和分镜描述输入 Kling、MiniMax 等视频模型，逐镜头生成。
5️⃣ 返工与剪辑： 用 FFmpeg 处理视频素材，再用剪映等工具完成剪辑、配音和字幕。人物崩掉就单独重做，不必整段推倒重来。
这套流程最麻烦的地方，是每一步都可能需要不同模型，还要反复生成。
账号、API、额度、账单散落在不同平台，工具切换本身就成了成本。
所以我越来越认同一个判断：
AI 创作的下一个效率提升点，不只是模型更强，而是把不同模型高效地串起来。
这也是我重新关注 ZenMux 的原因。
一个账户、一套 API，接入多家厂商的文本、图片和视频模型。One API. Whole workflow.
🎃 刚好，ZenMux 的万圣节活动 TRICK OR TOKENS? 还在进行：
• 每天现金 PAYG 消费满 $1，解锁一个模型盲盒
• 奖池包含 16 款模型，每天 3,000 份奖励
• 单份最高估算使用价值 $500（模型额度，非现金）
• 约 44% 概率开到 Claude / GPT / Gemini，具体以活动页为准
每个账户每天限 1 次，全程最多 5 次。每天 22:00（北京时间）刷新，活动截至 10 月 31 日 22:00。
注意：订阅用量、赠送额度和优惠券抵扣不计入这 $1。
如果你平时就在调 API、跑 Agent、做图或做视频，可以在正常使用时顺手参与，不必为了抽奖额外消费。
前往 ZenMux，查看并解锁今天的盲盒：
zenmux.ai/halloween
#ZenMux #AI #LLM #ModelRouting
