White balance is the process of calibrating a camera, image, or video to represent white objects as genuinely neutral: free from the colour cast introduced by the prevailing light source: ensuring that colours in the image accurately reflect the colours of the original scene rather than being skewed warm or cool by the temperature of the light. Different light sources emit light of different colour temperatures: direct sunlight at midday produces relatively neutral, white light; late afternoon sunlight is warm and amber-tinted; overcast sky light is cool and slightly blue; tungsten indoor lighting is distinctly warm and orange; fluorescent lighting tends toward green. Without white balance calibration, a camera sensor records these colour casts faithfully rather than correcting for them, producing images where white objects appear orange under tungsten light or blue in open shade, and where skin tones, fabrics, and environments carry colour distortions that conflict with how the human visual system naturally adapted to those same light sources in person.
Colour temperature ( the primary variable that white balance corrects for ) is measured in Kelvin (K). Lower Kelvin values (2,000–3,500K) represent warm, orange-tinted light such as candlelight and tungsten. Higher Kelvin values (5,500–7,000K) represent cool, blue-tinted light such as daylight and overcast sky. Camera white balance settings are expressed in Kelvin as well: setting the white balance to match the colour temperature of the prevailing light source neutralises its cast. Most cameras offer a set of preset white balance modes ( daylight, cloudy, shade, tungsten, fluorescent, flash ) that correspond to common lighting scenarios, as well as an auto white balance mode that analyses the image and attempts to neutralise casts automatically, and a manual or custom mode that allows the exact Kelvin value to be set or calibrated from a known neutral reference in the scene. For critical professional production, shooting in RAW format records the full sensor data without baking in the white balance, allowing it to be adjusted precisely in post-production without any quality loss: a significant advantage over adjusting white balance on rendered JPEG or video output, which has already discarded the raw sensor data.
In AI generation, white balance is rarely specified as a technical camera setting but is effectively communicated through descriptions of lighting colour and time of day that carry white balance implications. Specifying 'golden hour' lighting tells the model to produce warm, amber-tinted imagery consistent with late afternoon sunlight: the white balance equivalent of a camera set to around 3,500K. Specifying 'cool overcast light' produces imagery consistent with a camera set to 6,500K or higher. Precise colour consistency across a series of AI-generated clips: particularly important when intercut with each other or with live-action footage: is most reliably achieved through post-production colour correction that harmonises the white balance of all outputs rather than through prompt specification alone.
