Using All 14 AI Commands in the Salon
From /summary to /slide: the features you didn't know about
機械翻訳 / Machine-translated
2026年8月4日・8分で読める
The salon chat has AI built right in. It doesn't just answer questions — you can also use 14 commands. Since most people aren't aware of them, here's a complete rundown.
Just type a string starting with / in the chat input field. For commands that don't require arguments, you can simply tap the chip to send.
/helpDisplays the available commands and the current settings for that salon. When in doubt, start here.
/whoamiDisplays your username, plan, and role in the salon (owner / admin / member). Use this when you want to figure out why you can't perform a certain action.
/forget and /clearMakes the AI forget the conversation up to this point. After this, it will only reference content from here onward.
This is quite important. When the topic has shifted in a long thread but the AI keeps dragging in earlier context, inserting /forget brings its accuracy back. Both commands do the same thing.
/summarySummarizes the conversation in the thread so far. Great for catching up people who joined partway through. Even in a thread with 100 messages, one use of this gets you up to speed.
/membersDisplays a list of active members in the salon.
/search <query>Searches the web before responding. When you let the AI answer purely from its own knowledge, outdated information can slip in — use this command to get responses grounded in the latest information.
For questions like "What's the latest on ~?" or "What's the current price of ~?", don't ask plainly — add /search. The quality of the answer noticeably improves.
/image <instructions>Generates an image. Writing /image edit <instructions> lets you edit the most recently generated image.
Getting the ideal image on the first try is rare, so the standard workflow is: use /image to get a rough result, then refine it with /image edit.
/3dConverts the most recent image into a 3D model (GLB). No arguments needed.
One important note: you cannot create a 3D model directly from text. An image is always required first. The order is: create an image with /image, then run /3d.
/arCreates an AR post from the most recent 3D model and returns a URL. Opening that URL on a smartphone lets you place the model in real space and view it.
The post created is private by default — only people who have the URL can view it, making it ideal for sharing prototypes with your team.
/slide [topic]Based on the flow of the conversation and the given topic, the AI creates a slide deck and returns a URL.
You can specify the number of slides, the amount of information, and the style.
/slide 新機能の社内説明 pages=8 量=少なめ テイスト=ビジネス
Whatever you discussed in the chat becomes presentation materials, just like that. This is also private by default.
/slice and /slice-rejinSlices a 3D model into data for a 3D printer. /slice is for the standard FDM method, and /slice-rejin is for resin (SLA) printing.
Run /slice help for details. These two commands will be covered in a future article.
/code <what you want to build>Submits a development request to the code agent. Prior configuration on the salon side is required, so this is not available by default.
You can select the AI model to use from the salon settings.
Switching based on your use case is the practical approach. For casual chat, use a lighter model; for design discussions, use a heavier one. Adjust to match the nature of your salon.
You can turn off AI responses entirely. For salons focused on human-only conversation, or where AI messages become noise, just disable it in the settings.
Just because the commands exist doesn't mean you have to use them.
To wrap up, here are three combinations that actually work well.
/summary → write an article — Have the AI summarize what was discussed, then use that as the basis for an article./slide → share — Turn the conversation into presentation materials on the spot and show it to others./image → /3d → /ar — Turn an idea into an image, make it three-dimensional, and place it in the real world.The third one has a bit of a "secret trick" flow to it, so the next article will dedicate an entire post to walking through it.