I've been thinking about your point regarding credit consumption with these tools. I started exploring this at https://liveimage.ai (an AI video avatar generator) with a low-res preview system first. I'm curious if you think this kind of approach would be more useful - generating quick low-res previews that use minimal credits, then only processing high-res versions once you're happy with the result? Seems like it could help avoid wasting resources on full quality generations that don't match what you're looking for. I often wonder if jumping straight to high-res without previewing is partly why people burn through credits so quickly. Would love to hear if you've encountered other tools taking this kind of approach.