Accurate fish population monitoring is essential for sustainable aquaculture and marine ecology. Manual counting is labour-intensive, error-prone, and slow. FishTally automates species-specific counting by combining YOLOv8 for high-precision detection with ByteTrack for robust multi-object tracking. Its real-time detection–association pipeline maintains accuracy under occlusion, variable lighting, and water turbidity. A customisable command-line interface lets users tune thresholds and define line- or polygon-counting zones across diverse habitats. It processes continuous video streams for population analysis. Evaluations on DeepFish show strong precision and recall for single- and multi-species models, reducing human error, lowering costs, and improving time-sensitive decision-making for conservation and operations.