I’m making an avatar shopping game. Players can publish their outfits to a public feed, and other players can browse them, wear them, and buy the items.
Names and tags are already filtered with TextService. My two open problems are about the outfits themselves:
1. What counts as “bad”?
Every item is from the Roblox Marketplace, so each item is technically allowed. But a combination of allowed items can still make an outfit I don’t want kids scrolling past. Where do you draw the line? For example:
- Normal bikini / beach swimwear — fine or not?
- Very revealing / lingerie-looking outfits — I assume no, but where exactly does “revealing” start?
- Bodysuits or skin-tone clothing that looks nude in the thumbnail
- Shirtless classic torso
- “Sexy costume” stuff (maid, bunny) with normal coverage
- Gore / horror — normal Halloween seems fine, but how much blood is too much?
If your game has player-published content: what’s your list of “not allowed”, and how do you word it so it’s clear to players and mods?
Also — does a dataset or category list for this already exist anywhere? Like a shared taxonomy of “bad outfit” types, a reviewer rubric, or a blocklist of asset ids that are always trouble in combinations. To be clear, I mean labels and category definitions only, not actual content. I’d rather not invent this from scratch if someone has already done the work.
(Please just describe things in words — don’t post images of bad outfits here.)
2. What’s the best way to actually moderate it?
The options I see:
- Everything publishes instantly, players report bad ones, I review reports
- New players’ publishes wait for approval, trusted players go live instantly
- Everything waits for approval (probably too slow?)
- Auto-check somehow (asset blocklists, or screenshot the outfit and run it through an image moderation API)
What have you actually used, and what went wrong with it? Also, is there anything built into Roblox (AvatarEditorService, Open Cloud, anything) that helps with this beyond text filtering?
Thanks — real experience beats theory here, even from small games.