Language detection on a very large collection of documents can be done to increasing performance of information retrieval system. One of popular method on language detection is N-Grams, based on pieces of n-characters taken from a string. This research is developed language detection system based on N-Gram that performs by Indonesian or English language. In general, the steps being taken there were 3 phases, namely creating profile of each language, system testing, and system evaluation. Fifty documents were used to creating profile of each language, i.e. 25 Indonesian and 25 English. Sixty documents were used for system testing. System performance was evaluated using F-measures. Based on the test, obtained F-measures for unigram, bigram, and unigram respectively 0.933, 0.917, and 0.933.