Mixed ANOVA help?!

I have a 2x2 (Scene x Valence) ANOVA with Scene (Scene1, Scene2) being WS and Valence (Positive, Negative) being BS.
For each scene people read a description (which differs by valence) and answer DVs.

I did not want the Valence of Scene 1 to affect Scene 2 so for every person, if Scene 1 was Positive, Scene 2 was Negative and vice-versa, if Scene 1 was Negative, Scene 2 was Positive.

When I set this up on Qualtrics though, the BS variable was just "Valence" and it's set up so that:
Valence=1 is Positive Scene 1 and Negative Scene 2
Valence=2 is Negative Scene 1 and Positive Scene 2

The data output has the following columns: Participant ID, Valence, Scene1_DV, Scene2_DV.

Now I'm trying to run a RMANOVA with Scene as the WS and Valence as the BS and it makes zero sense because what Valence means is dependent on the scene.

If I want to look at how Scenes 1 and 2 differ within 1 level of Valence, then it seems not doable.

I think I have to make each scene its own row so that I can accurately represent the Valence for that one scene; then Participant ID is going to appear twice. Does someone know how to run a mixed ANOVA when the data is structured like this?