Bokeh (from the Japanese word 'boke', meaning blur or haze) is a visual style and photographic technique that focuses on the aesthetic quality of the out-of-focus areas of an image. In traditional photography and cinematography, it describes the way a lens renders points of light outside the depth of field, producing soft, creamy circles or shapes that add a sense of depth, isolation, and mood. In the context of Generative AI, bokeh has become a common prompt keyword and stylistic descriptor, used to instruct text-to-image and text-to-video models to produce backgrounds that are not just blurred but deliberately pleasing, often with distinctive highlights and a smooth, dreamy transition from sharp focus to soft defocus.
The term originated in Japanese photography circles in the late 20th century, popularized in the West by photographer Mike Johnston, who discussed it on his online forums in the 1990s. It was never a formal optical standard but rather a subjective appreciation of how a lens handles blur. Photographers like Ken Rockwell and Erwin Puts wrote about it extensively, grading lenses for their 'bokeh quality' - noting that some lenses produce harsh, double-edged, or busy blur while others deliver smooth, creamy, or silky backgrounds. This qualitative distinction carried over into film, where cinematographers deliberately use fast lenses, long focal lengths, and large apertures to isolate subjects, and into AI art, where the term now functions as a shorthand for a specific aesthetic.
Historical Origins in Photography
The concept of bokeh predates digital imaging, emerging from lens design and photographer discourse in Japan during the 1960s and 1970s. Japanese optical companies like Nikon and Canon produced lenses with complex glass elements, and enthusiasts began comparing how different designs rendered out-of-focus points of light. The term 'boke' was used in Japanese camera magazines such as 'Nippon Camera' and 'Asahi Camera' to describe the 'character' or 'flavor' of blur. Western photographers, particularly those who admired Japanese lens craftsmanship, adopted the term in the 1990s. Mike Johnston, founder of The Online Photographer, published an influential essay in 1997 titled 'The Bokeh Wars,' which helped standardize the spelling 'bokeh' to distinguish the aesthetic quality from mere blur.
Key optical factors influencing bokeh include the number and shape of aperture blades (circular bladed apertures produce rounder highlights), the lens's spherical aberration correction (under-corrected lenses often yield softer bokeh), and the distance between the subject and background. Historical lenses like the Minolta 58mm f/1.4 and the Zeiss Plana series are often cited for their pleasing bokeh, while some mirror lenses produce a notorious 'donut' bokeh with hollow centers. These characteristics became a selling point for 'fast glass' - lenses with large maximum apertures like f/1.2 or f/1.4.
Bokeh in Cinematography and Film
In film, bokeh achieved prominence starting in the 1970s with the work of cinematographers who embraced deep focus but also controlled shallow focus for dramatic effect. Directors like Stanley Kubrick used wide apertures in low-light scenes, as in 'Barry Lyndon' (1975), where candlelit shots exhibit a natural, painterly bokeh. More recently, the 'bokehrama' technique - shooting with a lens wide open and moving the camera - became popular for music videos and commercials. Notable practitioners include cinematographer Emmanuel Lubezki, known for his work on 'The Tree of Life' (2011) and 'Birdman' (2014), where he used natural light and AI-assisted post-processing to achieve dreamy, bokeh-heavy backgrounds. Another key figure is Roger Deakins, who often uses anamorphic lenses that produce oval-shaped bokeh highlights, adding a cinematic anamorphic flare. In the 2010s, digital cinema cameras and smartphone portrait modes simulated bokeh using computational photography, bringing the aesthetic to mass audiences. Films like 'Her' (2013) and 'La La Land' (2016) used bokeh as a visual metaphor for memory, longing, and romantic haze.
Visual Characteristics and Aesthetics
Bokeh is defined by specific visual markers. The most recognizable is the presence of smooth, round, or polygonal out-of-focus highlights - often light points like street lamps, Christmas lights, or water reflections - that appear as soft circles or 'bubbles.' The background transitions from sharp to progressively softer focus without harsh edges, creating a sense of dreaminess and separation. Typically, bokeh requires a shallow depth of field, achieved with a wide aperture (f/1.4 to f/2.8), a long focal length (85mm, 135mm, or 200mm on full-frame sensors), or a close subject distance. The quality of bokeh is often described as 'creamy,' 'silky,' 'smooth,' or 'organic' when it has no visible hard edges, versus 'busy' or 'nervous' when the blur pattern has double lines or swirling shapes. Some lenses produce 'cat's eye' bokeh, where highlights are oval or have pointed edges, while others like the Lomography Petzval lens create a swirling, vintage look. In AI-generated images, bokeh is typically invoked with keywords emphasizing 'background blur,' 'shallow depth of field,' 'bokeh lights,' or '50mm f/1.4' to replicate these optical effects.
Notable Practitioners and Artistic Use
Beyond photographers and filmmakers, fine artists and commercial photographers have used bokeh as a central compositional element. Japanese photographer Daido Moriyama often included grainy, high-contrast bokeh in his street scenes, while contemporary photographer Ryan McGinley uses it to capture nude subjects in lush, out-of-focus environments. In the digital art world, bokeh has been integrated into works by artists like Nathaniel Goldberg, whose fashion editorials rely on large-aperture portraits. In AI art, the aesthetic has become a staple because it immediately signals photographic realism and professional quality. AI models like earlier diffusion models and Midjourney have been trained on millions of photographs, allowing them to replicate bokeh effects when prompted. Community artists often use bokeh to add depth to portraits, product shots, and landscape scenes, sometimes overdoing it to create a hyper-real 'bokeh overload' that is popular in concept art and digital painting.
Bokeh in AI Image and Video Generation
In the realm of Generative AI, bokeh is a standard prompt element for several reasons. First, it helps ground the image in photographic realism, as real photos overwhelmingly have shallow depth of field. Second, it improves subject isolation, making the main object or person pop out from a busy background. Third, it aids in composition by providing a soft, non-distracting backdrop that guides the viewer's eye. Prompt writers use a variety of keywords: 'bokeh,' 'bokeh light,' 'cream bokeh,' 'circular bokeh,' 'soft bokeh,' 'dreamy bokeh,' 'background bokeh,' and technical camera settings like 'f/1.8,' '85mm,' 'portrait lens,' 'macro,' or 'defocused.' Negative prompts often include 'sharp background,' 'busy background,' or 'in focus' to avoid unwanted detail. Models like DALL-E 3, Claude (for text prompts describing scenes), and Imagen understand these terms, though results vary. In video generation, bokeh appears as a temporal effect - the blur stays consistent across frames, and moving highlights create light trails. Some models, such as Halcyon (a short-lived startup), explicitly marketed bokeh-heavy aesthetic generation.
Technical Considerations and Prompt Craft
Generating effective bokeh in AI output requires balancing several prompt components. The subject should be described as 'sharp' or 'in focus' to contrast with the blurry background. Terms like 'shallow depth of field' or 'isolated subject' reinforce this. For specific bokeh shapes, prompts can include 'hexagonal bokeh' or 'anamorphic bokeh' (oval, horizontal flare). The color and brightness of bokeh highlights can be influenced by describing 'golden bokeh lights,' 'blue bokeh circles,' or 'sparkling bokeh.' In diffusion models, increasing the prompt weight or using CFG (Classifier-Free Guidance) on 'bokeh' can intensify the effect, but over-weighting may lead to artifacts like black rings or unnatural blur. Some users add lens-specific terms like 'carl zeiss lens' or 'leica lens' as a proxy for high-quality bokeh. For video, adding 'camera breathing' or 'focus pull' can animate the bokeh. The use of Cross-Attention maps in models like Stable Diffusion allows users to manipulate which regions get the bokeh effect, though advanced tools are required.
Relationship to Other Visual Styles
Bokeh is closely related to but distinct from other blur-based styles. 'Tilt-shift' or 'miniature effect' uses selective blur to make scenes look like toys, often with a horizontal band of focus. 'Bokehrama' is a photographic technique where moving the camera during exposure creates bokeh streaks. 'Chromatic aberration' or 'fringing' produces red-green edge blur, which is a defect, not desired bokeh. In AI prompts, bokeh often pairs with 'cinematic,' 'photorealistic,' 'film grain,' 'anamorphic,' 'nostalgic,' or 'dreamlike.' It also overlaps with 'portrait photography,' 'macro photography,' and 'night photography' styles. Unlike 'depth of field' which is a technical measure, bokeh is an aesthetic judgment - you can have shallow depth of field with terrible bokeh, or deep focus with minimal blur. Prompt writers thus use 'bokeh' exclusively to enhance beauty, not merely to specify focus.
Future Directions and Tools
As Generative AI models improve, bokeh manipulation is becoming more granular. Tools like commercial image editors now use deep learning segmenters to apply bokeh selectively to generated or real images. Some prompt-based systems allow users to input 'focal length,' 'f-stop,' and 'distance' values numericallyholistically. Companies like Samsung and Apple use bokeh simulation in their smartphone camera modes, and AI models are trained on these outputs. However, AI-generated bokeh sometimes suffers from 'uncanny blur' - the background is too perfect, lacking the optical aberrations and sensor noise of real lenses. This has led to a trend of adding 'lens flare,' 'vignette,' or 'chemical film tones' to make bokeh feel authentic. Researchers at UC Berkeley and Stanford have published papers on depth-aware diffusion, which uses depth maps to naturally generate bokeh with correct occlusion and perspective. In 2024, OpenAI's Sora and Google's Veo demonstrated realistic video bokeh with temporal coherence, though access remains limited. The future likely sees bokeh as a user-controlled slider, similar to camera settings, rather than a vague prompt keyword.
Typical Prompt Examples
To illustrate, a common prompt might be: 'Portrait of a woman with blonde hair, sharp focus, soft bokeh background, golden hour light, bokeh lights like glitter, shallow depth of field, 85mm f/1.4, cinematic, photorealistic.' Another: 'A red sports car parked at night, with bokeh circles from street lights, background defocused, high contrast, film grain.' Negative prompts often include: 'sharp background, in focus, high details on background, flat depth.' The inclusion of 'bokeh' in a prompt often requires a trade-off: increasing background blur reduces background detail, which can be either desired or detrimental depending on the scene. Skilled prompt writers experiment with synonyms and camera terms to find the sweet spot between algorithmic interpretation and artistic vision.
Conclusion
Bokeh has evolved from a niche photographic term to a mainstream visual language. In AI generation, it serves as a bridge between optical physics and digital aesthetics, allowing users to emulate decades of cinematic and photographic technique without owning expensive lenses. Its versatility - applicable to portraits, landscapes, products, and abstract art - makes it a persistent keyword in the prompt engineering toolkit. As models become more physics-aware, the term will likely become more precise, but its core appeal remains: the transformation of distracting reality into a soft, dreamy, and emotionally resonant backdrop.