An RNN trained on 21 minutes of TikTok videos from the trend "Strawberries in Salt Water" and on scraped Tinder profile descriptions. The installation shows it reading: generation after generation, its output evolves from noise toward the voice of both corpora.
Process
Extraction. A Python script pulls profiles from Tinder and stores their fields as text: age, gender, distance, linked Instagram, job, school, number of pictures and the free-text description. The descriptions form the first corpus.
Second corpus. Twenty-one minutes of scraped TikTok videos from the viral Strawberries in Salt Water trend, transcribed to text. The same footage is the starting point of Average Clones or Strawberries in Salt Water.
Training. A recurrent neural network (RNN) is trained on both corpora to predict speech. The installation shows the model reading, generation after generation, as its output evolves.