3.2Composition, framing & lens
The essentials in 30 seconds
- Shot + angle = storytelling: close-up, medium, wide shot and low/high angle change the relationship to the subject — choose by intent, not at random.
- The lens is a shortcut: 35, 50, 85mm + aperture imply in one word a consistent geometry and blur — the antidote to the “AI” look.
- Format and text are planned: --ar at generation time (recropping destroys) and reserved negative space — text is added afterwards, never by the AI.
Running mini-case — Sonia prepares a shop banner in Algiers
Sonia, freelance graphic designer in Algiers, must deliver the web banner of a craft shop (16:9 format) with the brand name to add on the left. First generation: beautiful centered carpet, no free space, square format — unusable. She starts over by framing: “medium shot, carpet off-center to the right, negative space on the left, large empty wall, 85mm, shallow depth of field --ar 16:9”. The visual arrives in the right format with an empty area on the left: she sets the typography in five minutes in her graphics tool. Framing did the work, not retouching.
This same case returns in the lesson, the diagram, the script and the 48h challenge.
1. Shot, angle, composition: the narrative vocabulary
After light, composition and framing separate an amateur image from a mastered one. Shot size: extreme close-up (detail), close-up (face), medium shot (bust/waist), full body shot, wide shot/establishing shot (wide shot setting the scene). Angle: eye level (neutral), low angle (imposing), high angle (diminishing), bird's eye view/top-down, dutch angle (tension). Sonia picks the shot according to what she wants to tell: a craftsman at work in medium shot, his shop in wide shot.
Composition: rule of thirds, centered composition (symmetry), leading lines (lines guiding the eye), framing (frame within the frame), negative space (empty space — essential if text must be added). Placing the subject to one side mechanically frees the other side for typography.
2. The lens as a shortcut, the format from the start
The lens choice is a formidable shortcut: the model learned from millions of photos captioned with their technical data, so “85mm, f/1.8” activates in one go flattering tightness, plane compression and blurred background. Reference points: 35mm (wide, natural, reportage), 50mm (human vision), 85mm (portrait par excellence), telephoto (isolates), wide-angle (exaggerates, landscapes/interiors). Complete with aperture: f/1.8 or shallow depth of field (blurry background), f/11 or deep focus (everything sharp). One lens + one aperture are enough — never contradictory references.
Finally the format with --ar: 16:9 (banner/video), 9:16 (mobile/story), 1:1 (square post), 3:2 or 4:5 (photo). Picking the right ratio at generation time avoids recropping and losing quality — and text space. Professional rule: the image is generated, the text is added afterwards (Photoshop, Canva, Figma), where it stays editable and true to the brand guidelines.
🔎 Concrete example — Preparing for text addition at generation time
Sonia must reserve the left area of her banner for typography.
“handwoven carpet on the right, negative space on the left, large empty wall, 85mm, shallow depth of field --ar 16:9”
→ empty area for text, exact format of the medium, readable blurry background.
→ text added afterwards in the graphics tool.Reading: a dark, plain area welcomes light text with no retouching.
Lesson: never ask the AI to write the text (mistakes, uncontrolled fonts).
Section vocabulary
- Shot size
- Apparent distance to the subject: close-up, medium shot, full body, wide shot / establishing shot.
Simple example: Sonia's craftsman in medium shot, his shop in wide shot.
What it is not: framing left to the model's chance.
Reflex: pick the shot by narrative intent.
- Camera angle
- Camera position: eye level, low angle, high angle, top-down, dutch angle.
Simple example: The low angle making the pottery monumental.
What it is not: an angle contradicting the message (high angle on a hero).
Reflex: low angle to impose, high angle to diminish.
- Composition rules
- Rule of thirds, centered composition, leading lines, framing, negative space (free space for text).
Simple example: The off-center carpet freeing the left side for the shop name.
What it is not: a centered subject leaving no room for text.
Reflex: go off-center as soon as text will come.
- Focal length (lens)
- 35mm (wide/natural), 50mm (human vision), 85mm (portrait), telephoto (isolates), wide-angle (exaggerates).
Simple example: The 85mm for the craftswoman portrait in Algiers.
What it is not: stacking “24mm, telephoto, macro”.
Reflex: one lens + one aperture per prompt.
- --ar (aspect ratio)
- Parameter setting the image format (16:9, 9:16, 1:1, 4:5…); to pick at generation time.
Simple example: Sonia's exact 16:9 banner.
What it is not: generating square then recropping to a banner.
Reflex: start from the final medium's format.
In practice — Composing and framing deliberately
- Take one subject and generate it in turn as close-up, medium shot then wide shot: measure each shot's narrative effect.
- Test angles: low angle then high angle on the same subject, and observe the shift in power relationship.
- Add a lens (35mm, then 85mm) and a depth of field (shallow depth of field): compare subject/background separation.
- Generate a version with “negative space” and the right --ar for your medium (banner, story, square post).
Ready-to-say script — explaining your framing to the client (30 s)
“I framed for your use: subject off-center to the right, empty space on the left for your name, exact 16:9 format of your site, blurry background for readability. I add the typography in the brand guidelines after generation — so the text stays editable for your variants.”
To say on delivery; show the visual before/after adding the text.
Copy-paste template — framing prompts
artisan weaving a carpet, medium shot, eye level, subject on the right, negative space on the left, 85mm f/1.8, shallow depth of field --ar 16:9
pottery jars in a souk, low angle, wide shot, leading lines of the alley, golden hour --ar 3:2
pastry box on marble, top-down view, centered composition, minimal background with copy space --ar 1:1Freelancer
One shot + one angle + the right --ar: the trio making each client visual directly usable.
SME in Algiers
Sonia delivers the banner in the exact format, text area reserved, typography set in five minutes.
Project team
Impose the medium's --ar and negative space in every visual brief before generating.
Beginner
This week: one subject in three shots, two angles compared, one version with negative space.
✓ Best practices
- Pick shot and angle by narrative intent, not by default.
- Use the lens as a shortcut: one lens + one aperture are enough.
- Set --ar and plan negative space at generation time if text will come.
✗ Common pitfalls
- Recropping afterwards: lost quality and destroyed text space.
- Stacking contradictory lens references (“24mm, telephoto, macro”).
- Forgetting negative space when text must be added — or letting the AI write the text.
Key takeaways
- Shot vocabulary (close-up → wide shot) and angle vocabulary (low/high angle, top-down): immediate narrative levers.
- Composition: rule of thirds, leading lines, framing, negative space (essential if text must be added).
- Lens (35/50/85mm, telephoto, wide-angle) + aperture imply realistic geometry and blur.
- Pick the format with --ar at generation time rather than recropping afterwards (quality loss).
48h challenge — apply tomorrow
Deliver 1 client visual with the right --ar with reserved negative space + 1 comparison 35mm vs 85mm on the same subject.
- I pick shot and angle by intent.
- I name one lens + one aperture per photo prompt.
- My format (--ar) and text area are planned from generation.
Frequently asked questions
Why specify a lens when the AI has no camera?
Because the model learned from millions of photos captioned with their technical data. Writing “85mm, f/1.8” activates a whole set of associated visual characteristics: tightness, plane compression, blur of a certain texture, typical distance. It is a remarkable shortcut: one term replaces three approximate sentences and brings consistency that is hard to get otherwise — credibility and internal consistency. Pick the lens by intent (wide-angle to immerse, 85mm/telephoto to isolate) and do not overload. This trick applies mostly to photographic rendering; for illustration or 3D (3.4), it is another vocabulary.
How do I get an image where I can cleanly add text?
It is prepared AT generation time. The key term is NEGATIVE SPACE: ask for it and specify where (“negative space on the left”, “copy space”). Steer the composition (off-center subject), pick the right --ar from the start, favor a simple or blurry background, and think CONTRAST (light text → dark area). Important point: do NOT ask the AI to write the text — typographic rendering stays random. Constant practice: generate the image, add the text afterwards (Photoshop, Canva, Figma) — the combined-workflow logic (4.5).
What should you ask for to reserve an area for text in the image?