The Frog Prince

I finished my first entry for the CGHub challenge "Fairy Tales reinvented", that quickly became an entry for the Art Order challenge as well, with Schindehette's permission.
I chose the Frog Prince tale, but I didn't really reinvent it, I just once again focused on little symbols and hidden meanings, naughty humour and such. Nothing that the tale itself doesn't already implicitly represent.
That's why I think it suits the ArtOrder challenge much more than its CGHub counterpart and I'm now drawing an Alice in Wonderland reinterpretation for that.

Anyway, here's the Frog Prince.

Original height is a little over 6000 pixels. The entire illustration was fairly straightforward and smooth except for the girl. I redrew her completely at least 3 times, due to anatomical problems and such. As usual I never start with reference. I adopt it through the process if necessary.
Since I have the "bad" habit of picking weird angles I had to use some reference for the head and shoulders, but had to make the rest up.
Hands are based on reference too (my own hands) for realism and style purposes.

Since I start moving more and more towards realism I notice my need for reference is gradually increasing, but I will never start anything with reference. Whenever I did that I noticed it killed my vision of the piece, it constrained me to what I was seeing in the reference, becoming a true obstacle.
I also don't particularly like how something entirely based on reference looks like. It feels like it's not my own work, it feels alien and depressing.
The least reference I use the best it is for me. These are personal choices and I disagree with those who promote using reference whenever we encounter a little problem.
Our memory is filled with precise images of the entire world, we just don't have access to it with great lucidity, but it's there. We can train to access it and to refresh those memories though.
First I try to figure out what I've done right. Once I know why I've failed on what looks bad I look for help (opinions, reference).
If we don't give ourselves the chance of failing through research we'll always be reference-slaves. It's good to research and experiment when doing personal work or contests with generous deadlines (most are).