corrected the voice conversion note, melody behaviour is not what the card said

README.md

@@ -105,8 +105,12 @@ as a suggestion and tends to read straight past anything outside the common set.
 
 The runtime can also respeak an existing recording in a different voice, keeping the original timing,
 phrasing and emphasis while changing only the speaker. This is not part of the upstream model, it falls
-out of how the codec separates semantic content from acoustic detail. It is experimental, and pitch does
-not carry over, so singing comes out spoken unless you retain some source acoustic codebooks.
+out of how the codec separates semantic content from acoustic detail.
+
+It is experimental. Pitch is regenerated rather than copied, so a converted vocal is re-sung in the
+target voice's own register instead of at the source's. Whether a tune survives varies clip to clip,
+and `keep_acoustic 1` or `2` copies the lowest acoustic codebooks from the source to pull more of the
+original contour through. Judge it by ear, and leave `keep_acoustic` at 0 for ordinary speech.
 
 ## Conversion