card: describe the encoding instead of linking the recipe

README.md

@@ -107,5 +107,6 @@ transcript.
 ## Licence and attribution
 
 The weights are a derivative of NVIDIA's parakeet-tdt-0.6b-v3 re-trained by Fermion Research and are distributed under
-CC-BY-4.0, as upstream. `NOTICE` reproduces Fermion Research's change list and training-data attribution. The Core ML
-conversion recipe is in [mobius](https://github.com/FluidInference/mobius) (`models/stt/phonon-2/coreml`).
+CC-BY-4.0, as upstream. `NOTICE` reproduces Fermion Research's change list and training-data attribution. The encoders are
+exact re-encodings of the upstream weights (grouped-channel palettes, or a sparsity mask plus palettes over the non-zeros);
+the decoder and joint are the upstream int6 tables in fp16.