V1 good for anime only, with success rate about 25-50%. The test tubes come out on the big side, but maybe careful prompting can fix that.

V2 is more flexible and easier to use with better tags for size, position, etc. I spent a lot more time tagging and also cleaning up the reference images (removing extra fingers etc). Should work with anime, comic book, 3d render and photo styles.

TAGS EXPLAINED:

testtube OR testtubes -- you want 1 or multiple

TTtiny OR TTsmall OR TTmedium OR TTlarge -- size

TTliquid-clear OR TTliquid-red/pink/orange/yellow/blue/green -- colour of liquid

TTdrinking -- drinking

TTlying -- lying on side

TTspill -- spilling out

TTfull OR TThalf-full OR TTnear-empty OR TTempty -- how much liquid

TTclosed OR TTopen -- is there a cork or not?

TTgrab-bottom, TTgrab-top, TTpinch-bottom, TT-pinch-top, TTpinch-all -- possible hand grabbing positions
TTbubbles -- bubbly liquid?

TTrack -- if you want lots of tubes in a test tube rack

Include "hand, holding" if you want a hand in the shot.

Include "still life, no humans" if you don't want a hand or any human.

DO NOT include "potion" as it tends to create a circular-bottomed flask instead of a test tube.