This study investigates grapheme encoding in Croatian as a second language among Farsi speakers after twenty hours of learning. Three phases of encoding tasks were administered: 1) dictation of individual phonemes, 2) dictation of words beginning with those phonemes, and 3) dictation of simple sentences with words from the previous phase. Respondents used "-" to denote unencoded items. Eleven Afghan respondents at the beginner level (A1 according to CEFRL), aged 18 to 63, were sampled conveniently. The study aims to assess: a) accuracy in encoding individual graphemes and words, b) problematic graphemes, and c) accuracy in encoding complete words. The results will illuminate initial decoding specifics for this group, confronting the added complexity of differing graphic systems between L1 and L2. Furthermore, implications for Croatian orthography acquisition as L2 will be discussed. Analysis of encoding by Farsi-speaking Croatian learners showed overall success with sentences but difficulty with individual graphemes, possibly due to reliance on lexical rather than phonological knowledge. Notably, struggles were observed with "nj," unlike with "c" as seen previously. This might be because "nj" is less common in Croatian, especially early on. Transfer errors from Farsi, like omitting short vowels, were evident. Instruction for Farsi learners should focus on specific grapheme errors such as "nj," "ć," "dž," and "đ," as well as consonant clusters and short vowel encoding in Croatian.