Fast H.264 transcoding

I am working on (sort of) a VOIP app for mobile with video support. Most modern smartphones are equipped with a hardware H.264 encoder / decoder. However, most of them only support a limited number of encoding profiles. In particular, I am interested in SVC, which is usually not supported. The simplest solution is to use a software encoder. But, obviously, it has its flaw.

I recently got the idea to create a hybrid hardware + software encoder. Which consists of a hardware encoder and a "small software editor".

I know (in general terms) that H.264 works in terms of macroblocks. The coded stream contains instructions for macroblocks: either a full retransmission (in the form of jpeg-like data), its motion parameters, or just nothing.

A complete software transcoding will process such a coded stream, parse / implement it and execute on its graphical data, and then encode the resulting image, which includes processing each macroblock, deciding on a strategy, generating motion vectors and finally encoding all instructions.

The idea is to do a "shallow" conversion. I'm, however, about parsing the incoming codestream, but instead of executing its instructions - just use them as a source for the encoder, that is, encode them in a complex H.264 encoding scheme with the required accuracy (that is, bitrate), etc., Without actually analyzing the image, finding motion vectors, etc.

Basically it uses decoder + encoder but with a shortcut so that the actual image is not generated by the decoder and not parsed by the encoder.

Is it possible? Was there an attempt to do something similar?