Maybe I'm stating things you already know, but in case you don't:
It sounds like you want to change the dynamics of the mix. Bring up the quiet ambient sounds, push down the dialogue, right? So it is impossible to obtain the dialogue and ambience as separate tracks?
The tool for this is a compressor with little character (so very transparent), preferably a multi-band compressor or at least one with side-chain eq.
Set the threshold so that it triggers on the dialogue and not on the ambience (if you have it, the sidechain-eq set to the most prominent frequency area of the vocal can help).
Then you can bring up everything so that the vocals are at the same level as before, but the ambience is now louder.
Tweak the ratio, threshold and release while listening for undesired pumping effects at points where someone starts or stops talking.
With extreme settings this will make it very tiring to listen to. If the difference in dynamics is very big you will have to push the compressor very hard which will eventually make it a mushy mess.
Another method that you could try if the above doesn't work is just automating the level by hand. Very time-consuming, but you have the most control.