{"course":{"slug":"ai-images","title":"Image Generation, in Practice","url":"https://addaly.com/learn/ai-images","level":"beginner","track":"create"},"lesson":{"slug":"what-to-shoot-instead","title":"The frames where a phone camera wins","url":"https://addaly.com/learn/ai-images/what-to-shoot-instead","position":19,"of":69,"minutes":7,"answer":"Anything that must be exactly itself (a real product, real premises, real staff, food you are selling) is faster, cheaper and safer to photograph than to generate.","keyIdea":"Anything that must be exactly itself (a real product, real premises, real staff, food you are selling) is faster, cheaper and safer to photograph than to generate, and recognising those frames early is a professional skill rather than a failure."},"body":{"format":"markdown","text":"## The unfashionable answer\n\nSome of the frames on your brief should not be generated at all. Saying so early makes you more useful, not less, and it saves the day that would otherwise be spent discovering it.\n\nThe test is simple: **does this image make a claim about a specific real thing?** If yes, generating it is either impossible, dishonest, or more expensive than photographing it.\n\n## The categories\n\n**The actual product.** A generated bottle is a bottle-shaped object with an unreadable label, wrong proportions and a logo that is nearly right. Getting the real one is thirty minutes: a window, a white card as a reflector, a sheet of paper as a background, and a phone on a stack of books. Then generate the *environment* and composite the real product into it, which is a recognised professional workflow and covered later in this course.\n\n**Food you are selling.** Food photography is regulated in many markets to the extent that the image must represent the actual item. Beyond the law, generated food has a specific failure: it looks delicious and wrong, with impossible textures and garnish that does not correspond to anything. A restaurant that posts generated pictures of its dishes is setting up every customer for disappointment.\n\n**Your premises, your staff, your equipment.** These are claims about reality. A generated \"our team\" photograph is a fabrication in a way that a generated abstract background is not, and it is the single fastest way to lose the trust the image was meant to build.\n\n**Anything a customer will compare against the real item.** Clothing, furniture, hardware. If the generated version is prettier than what arrives, you have manufactured a complaint.\n\n**Documentary or journalistic content.** No exceptions worth discussing. The later block on rights and disclosure covers this properly.\n\n## What generation is better at\n\nThe list is not one-sided, and the frames below are ones where photography would be slower, costlier or impossible:\n\n- **Environments and backgrounds** that do not need to be a specific place.\n- **Concepts and abstractions**: growth, connection, an idea in a deck.\n- **Things that do not exist**: a product concept, a fantasy scene, an illustration.\n- **Sets you cannot afford**: a mountain road at dawn, a 1970s office, a crowded market at blue hour.\n- **Texture, pattern, surface** for backgrounds and overlays.\n- **Variations for testing**: twelve versions of a banner to see which performs.\n\nThe sharpest formulation is this: **generate the world, photograph the claim.**\n\n## The phone is enough\n\nThe objection is always that there is no camera and no studio. A modern phone in daylight next to a window is a competent product camera, and the gap between a phone photograph and a studio one is mostly lighting, which costs nothing:\n\n1. A window, subject side-on to it, not facing it.\n2. A white A4 sheet on the shadow side, angled to bounce light back.\n3. A plain background: a sheet of paper, a wall, a piece of card.\n4. Phone on something solid. Tap to focus. Turn the flash off.\n5. Take thirty and keep two.\n\nThen clean it up. Background removal, colour correction and cut-out are exactly what the image-editing course covers, and all of it is available in GIMP and Photopea without spending anything.\n\n## Saying it to a client\n\n\"That one we should photograph: it will look better and it will be faster\" is a sentence that builds trust, because it demonstrates that you are choosing the tool rather than defending it. The version that damages you is the one where you generate it anyway, they notice the label is wrong, and you explain afterwards that the model cannot do labels.\n\nThe more useful frame for the conversation is that this is normal creative practice. Nobody illustrates a product catalogue; nobody photographs a concept for an annual report. Choosing per frame is the job, and it has always been the job.\n\n## The middle category\n\nBetween \"photograph it\" and \"generate it\" sits a third answer that handles a surprising share of briefs: **photograph the claim, generate everything around it**.\n\nThe bottle is real and shot against a white sheet. The mountain road it is standing on does not exist. The composite is honest because the product is what it is, and it is affordable because nobody drove to a mountain.\n\nThe same logic applies widely. A real headshot placed into a generated environment. A real building photographed on a dull day with a generated sky, which retouchers have done for thirty years without anyone calling it deception. A real dish on a generated table.\n\nWhere the line sits is a judgement, and it is worth stating the principle rather than a rule: **the parts of the image a viewer would rely on should be real, and the parts that are stage dressing need not be.** A sky is stage dressing. A dish, in an advertisement for that dish, is not.\n\nWhen you are unsure, the test that has never failed anybody is to imagine explaining the composite to the person it depicts, or to the customer who buys on the strength of it. If that explanation is comfortable, the composite is fine. If it is not, you already know."},"check":{"question":"What distinguishes the frames that should be photographed from the ones that should be generated?","options":[{"text":"Whether the subject contains fine detail such as text or a logo, since those are the specific elements that current generative models reproduce least reliably in practice.","correct":false,"tempting":"Names a real weakness but the wrong criterion: a generated abstract background with no text at all is still fine, while a plain unbranded product still needs to be the actual one."},{"text":"Whether the image asserts something about a specific real thing that a viewer could check against reality.","correct":true},{"text":"Whether photographing it would be cheaper than the compute and time that generating an acceptable version of the same frame would consume.","correct":false,"tempting":"Cost is a secondary consideration; a cheap generated staff photo is still a fabrication, and an expensive photographed environment may still be the wrong call."},{"text":"Whether the final image will be seen at a large enough size for generated artefacts and inconsistencies to become visible to an ordinary viewer.","correct":false,"tempting":"Display size affects how much repair a frame needs, not whether the image is making a factual claim that has to be true."}]},"prev":{"slug":"one-image-or-a-set","title":"One hero or twelve pieces is a different job","url":"https://addaly.com/learn/ai-images/one-image-or-a-set"},"next":{"slug":"writing-the-spec","title":"The one page that ends the revision argument","url":"https://addaly.com/learn/ai-images/writing-the-spec"},"authorship":"Founder and series editor: Dr Mustafa Mun. Written by AI models to a written standard. Exam keys checked blind by a second model: 3,665 of 3,666 agreed. Academic review invited.","review":{"status":"Draft, open for academic review"},"cite":"Addaly (2026) 'The frames where a phone camera wins', in Image Generation, in Practice. Edited by M. Mun. Addaly. Available at: https://addaly.com/learn/ai-images/what-to-shoot-instead (Accessed: 13 October 2026).","licence":{"name":"CC BY-NC-SA 4.0","url":"https://creativecommons.org/licenses/by-nc-sa/4.0/"},"reader":{"schema":1,"title":"The frames where a phone camera wins","courseTitle":"Image Generation, in Practice","module":{"index":2,"slug":"the-brief","title":"From a request to a specification you can generate against"},"answer":null,"keyIdea":"Anything that must be exactly itself (a real product, real premises, real staff, food you are selling) is faster, cheaper and safer to photograph than to generate, and recognising those frames early is a professional skill rather than a failure.","hero":{"kind":"plate","image":null,"alt":null,"caption":null},"blocks":[{"t":"h2","id":"the-unfashionable-answer","runs":[{"text":"The unfashionable answer"}]},{"t":"p","runs":[{"text":"Some of the frames on your brief should not be generated at all. Saying so early makes you more useful, not less, and it saves the day that would otherwise be spent discovering it."}]},{"t":"p","runs":[{"text":"The test is simple: "},{"text":"does this image make a claim about a specific real thing?","b":true},{"text":" If yes, generating it is either impossible, dishonest, or more expensive than photographing it."}]},{"t":"h2","id":"the-categories","runs":[{"text":"The categories"}]},{"t":"p","runs":[{"text":"The actual product.","b":true},{"text":" A generated bottle is a bottle-shaped object with an unreadable label, wrong proportions and a logo that is nearly right. Getting the real one is thirty minutes: a window, a white card as a reflector, a sheet of paper as a background, and a phone on a stack of books. Then generate the "},{"text":"environment","i":true},{"text":" and composite the real product into it, which is a recognised professional workflow and covered later in this course."}]},{"t":"p","runs":[{"text":"Food you are selling.","b":true},{"text":" Food photography is regulated in many markets to the extent that the image must represent the actual item. Beyond the law, generated food has a specific failure: it looks delicious and wrong, with impossible textures and garnish that does not correspond to anything. A restaurant that posts generated pictures of its dishes is setting up every customer for disappointment."}]},{"t":"p","runs":[{"text":"Your premises, your staff, your equipment.","b":true},{"text":" These are claims about reality. A generated “our team” photograph is a fabrication in a way that a generated abstract background is not, and it is the single fastest way to lose the trust the image was meant to build."}]},{"t":"p","runs":[{"text":"Anything a customer will compare against the real item.","b":true},{"text":" Clothing, furniture, hardware. If the generated version is prettier than what arrives, you have manufactured a complaint."}]},{"t":"p","runs":[{"text":"Documentary or journalistic content.","b":true},{"text":" No exceptions worth discussing. The later block on rights and disclosure covers this properly."}]},{"t":"h2","id":"what-generation-is-better-at","runs":[{"text":"What generation is better at"}]},{"t":"p","runs":[{"text":"The list is not one-sided, and the frames below are ones where photography would be slower, costlier or impossible:"}]},{"t":"ul","items":[[{"text":"Environments and backgrounds","b":true},{"text":" that do not need to be a specific place."}],[{"text":"Concepts and abstractions","b":true},{"text":": growth, connection, an idea in a deck."}],[{"text":"Things that do not exist","b":true},{"text":": a product concept, a fantasy scene, an illustration."}],[{"text":"Sets you cannot afford","b":true},{"text":": a mountain road at dawn, a 1970s office, a crowded market at blue hour."}],[{"text":"Texture, pattern, surface","b":true},{"text":" for backgrounds and overlays."}],[{"text":"Variations for testing","b":true},{"text":": twelve versions of a banner to see which performs."}]]},{"t":"p","runs":[{"text":"The sharpest formulation is this: "},{"text":"generate the world, photograph the claim.","b":true}]},{"t":"h2","id":"the-phone-is-enough","runs":[{"text":"The phone is enough"}]},{"t":"p","runs":[{"text":"The objection is always that there is no camera and no studio. A modern phone in daylight next to a window is a competent product camera, and the gap between a phone photograph and a studio one is mostly lighting, which costs nothing:"}]},{"t":"ol","start":1,"items":[[{"text":"A window, subject side-on to it, not facing it."}],[{"text":"A white A4 sheet on the shadow side, angled to bounce light back."}],[{"text":"A plain background: a sheet of paper, a wall, a piece of card."}],[{"text":"Phone on something solid. Tap to focus. Turn the flash off."}],[{"text":"Take thirty and keep two."}]]},{"t":"p","runs":[{"text":"Then clean it up. Background removal, colour correction and cut-out are exactly what the image-editing course covers, and all of it is available in GIMP and Photopea without spending anything."}]},{"t":"h2","id":"saying-it-to-a-client","runs":[{"text":"Saying it to a client"}]},{"t":"p","runs":[{"text":"“That one we should photograph: it will look better and it will be faster” is a sentence that builds trust, because it demonstrates that you are choosing the tool rather than defending it. The version that damages you is the one where you generate it anyway, they notice the label is wrong, and you explain afterwards that the model cannot do labels."}]},{"t":"p","runs":[{"text":"The more useful frame for the conversation is that this is normal creative practice. Nobody illustrates a product catalogue; nobody photographs a concept for an annual report. Choosing per frame is the job, and it has always been the job."}]},{"t":"h2","id":"the-middle-category","runs":[{"text":"The middle category"}]},{"t":"p","runs":[{"text":"Between “photograph it” and “generate it” sits a third answer that handles a surprising share of briefs: "},{"text":"photograph the claim, generate everything around it","b":true},{"text":"."}]},{"t":"p","runs":[{"text":"The bottle is real and shot against a white sheet. The mountain road it is standing on does not exist. The composite is honest because the product is what it is, and it is affordable because nobody drove to a mountain."}]},{"t":"p","runs":[{"text":"The same logic applies widely. A real headshot placed into a generated environment. A real building photographed on a dull day with a generated sky, which retouchers have done for thirty years without anyone calling it deception. A real dish on a generated table."}]},{"t":"p","runs":[{"text":"Where the line sits is a judgement, and it is worth stating the principle rather than a rule: "},{"text":"the parts of the image a viewer would rely on should be real, and the parts that are stage dressing need not be.","b":true},{"text":" A sky is stage dressing. A dish, in an advertisement for that dish, is not."}]},{"t":"p","runs":[{"text":"When you are unsure, the test that has never failed anybody is to imagine explaining the composite to the person it depicts, or to the customer who buys on the strength of it. If that explanation is comfortable, the composite is fine. If it is not, you already know."}]}],"check":{"question":[{"text":"What distinguishes the frames that should be photographed from the ones that should be generated?"}],"options":[{"text":[{"text":"Whether the subject contains fine detail such as text or a logo, since those are the specific elements that current generative models reproduce least reliably in practice."}],"correct":false,"tempting":[{"text":"Names a real weakness but the wrong criterion: a generated abstract background with no text at all is still fine, while a plain unbranded product still needs to be the actual one."}]},{"text":[{"text":"Whether the image asserts something about a specific real thing that a viewer could check against reality."}],"correct":true,"tempting":null},{"text":[{"text":"Whether photographing it would be cheaper than the compute and time that generating an acceptable version of the same frame would consume."}],"correct":false,"tempting":[{"text":"Cost is a secondary consideration; a cheap generated staff photo is still a fabrication, and an expensive photographed environment may still be the wrong call."}]},{"text":[{"text":"Whether the final image will be seen at a large enough size for generated artefacts and inconsistencies to become visible to an ordinary viewer."}],"correct":false,"tempting":[{"text":"Display size affects how much repair a frame needs, not whether the image is making a factual claim that has to be true."}]}],"order":[2,3,0,1],"why":null},"prev":{"slug":"one-image-or-a-set","title":"One hero or twelve pieces is a different job","minutes":7},"next":{"slug":"writing-the-spec","title":"The one page that ends the revision argument","minutes":7}}}