The mystery of Merlin’s Charm of Making is, alas, no longer a mystery. Although Merlin and Morgana both pronounce things differently from each other, and even Merlin has two sounds which to me sound like phonemes but which must be allophonic, I get the following from the Charm of Making in John Boorman’s film Excalibur:

/a'na:l naθ'rax, u:rθ va:s be'θud, dox'je:l 'djenve:/

It’s certainly not Welsh. It looks very much like an attempt at Old Irish. (One wonders where Boorman got it.) Following is the best I can do at reconstructing reasonable Old Irish from it. It is probably a defective reconstruction. I have normalized to Modern Irish orthography to indicate lenition.

In Old Irish
Anál nathrach, orth’ bháis’s bethad, do chél dénmha

In Modern Irish:
Anáil nathrach, ortha bháis is beatha, do chéal déanaimh

In English:
Serpent's breath, charm of death and life, thy omen of making.

anál nathrach = breath of serpent
orth’ bháis ’s bethad = spell of death and of life
do chél dénmha = thy omen of making

anál fem. -á stem ‘breath, breathing’
nathair fem. -k stem ‘snake, serpent’ g. sg. nathrach
ortha fem. -n stem ‘prayer; incantation, spell’, from Latin oratio
bás masc. -o stem ‘death’ g. sg. báis
ocus conj. ‘and’ here shortened to 's
betha masc. -t stem ‘life’ g.sg. bethad
do prn. ‘thy’ Usually unstressed
cél masc. -u stem ‘omen, augury, portent’
dénumh masc. -m stem ‘making, doing’ g.sg. dénmha

Modern Irish would have the -is in bháis as a /sh/ sound, but it might not have been so palatalized in the Old Irish period; and the nonpalatal ’s of ‘and’ ought to reinforce that. The third part of the charm could also be dochél dénmha ‘an evil omen of making’, but that suits the sense badly. The word do ‘thy’ is usually unstressed in speech but what can you do...

Note that Merlin says dénmhe, which ought to be dénmha; perhaps there is some sort of ‘incantation register’ in which a final vowel can be altered in this way.... In any case, I am less than happy with the third part of this. I'd like to have seen an imperative or hortative, but verb-first syntax precludes even dénae, the imperative of do-gní (from which the verbal noun dénumh is formed), which anyway doesn’t have the nominal formative -mh.

I would be interested in hearing from specialists in Old Irish as to their opinions of this. There are other possibilities for the retro-translation, and indeed the use of a Latin loanword, given the context, is problematic.

Fuente: https://www.evertype.com/misc/charm.html