Hello Hans,
best pack the img in a box and then put some qQ around it with the clip at the end (the usual experimenting with origin), so basically something compatible to
\pdfliteral {q clip spec} \pdfximage{...} \pdfliteral {Q}
ending up in something
that works fine, thank you very much! This is my test code now: ----------------------- \directlua{ a=node.new("whatsit","pdf_literal") b=node.new("whatsit","pdf_literal") a.data = "q 0 w 20 20 m 140 20 l 140 160 l 20 160 l W n " b.data = "Q" image = img.scan{filename="cow.pdf"} i = img.node(image) node.insert_after(a,a,i) node.insert_after(a,i,b) head = node.hpack(a) tex.box[0] = head } \box0 \bye ----------------------- I wonder how it compares to the slightly different approach pdflatex uses (pdftex.def?). If I see it correctly the graphicx package creates an xobject/form from the bitmap and changes the bounding box there. Thanks again! Patrick