the built in filter just makes nsfw content choppy. So if making such content then, well first find a more visited lemmy thread/post on the subject of the text to image generator/&or the new model, hopefully you run into one with my JSON/&Dev prompting notes gemini convo, in that convo that or if you reactivate the convo after opening it you should be able to figure out the "string" & or prompt command to true/false such contents. Second the model in‐line filter will only make the generation of such things a bit less choppy if turned off, as it isn't a full block "pretty much no filter is a full block".