ai-product-photography
The video argues that accurate small-print typography on AI-generated product visuals comes from controlling the structure of the prompt (the order and priority of information) and from reinforcing product identity with reference images. Fixing text afterwards is not the approach. It claims the same hierarchy carries from still images in Nano Banana Pro to video in Kling 3.0 Omni.
The video states the problem: AI makes product images easy, but small branding text often gets distorted, slightly redesigned or misspelled. It proposes controlling typography from the start of generation, not fixing it later.
The central claim is that the lever is structure, not complexity or length. What matters is the order and priority of information in the prompt.
The video argues that models improvise by default. Each layer of the prompt exists to remove a freedom the model would otherwise use to drift, such as redesigning the product, reinterpreting colors, or inventing branding. This is said to reduce correction cycles and protect brand integrity.
Typography is treated as a first-class constraint, not a detail inside the product description. The video presents it as one of the most important parts of the prompt.
The evidence offered is a running shoe with technical text on the heel and mid-sole, shown in two different still-life environments (sandstone and terracotta). Zoomed in, the text stays sharp, aligned, readable and faithful to the real product. A shortened prompt with the same hierarchy produced consistent results, which the video takes as proof that order beats length.
The video then argues that if the structure is solid, typography should survive motion. In video, identity is carried by continuously active image references, not by re-describing typography in the prompt, so the lettering stays readable across camera moves and close-ups.
The video concludes that perfect typography is the result of structure plus reinforcement, not luck. A clear prompt hierarchy makes iteration shorter, corrections minimal and brand consistency predictable.
Structured prompt hierarchy (the seven-part formula) — An ordered prompt structure of composition, exact identity, typography control, true color, lighting harmony, optical clarity and negative constraints, which the video says works because of order and priority, not length. Apply: Write the prompt in that fixed sequence. The video says a shorter prompt that keeps the same hierarchy still gives consistent results.
Model selection: Google Nano Banana Pro for product typography — The video picks Nano Banana Pro in the image generator because it handles fine product details and small typography more reliably than most when working with realistic product photography. Apply: Choose Google Nano Banana Pro in the model selector before starting still-life product generation.
High-quality multi-angle studio references — Uploading several sharp, cleanly lit studio images of the product as references, since weak source images make typography accuracy harder to control. Apply: Upload multiple angles with clean lighting and sharp detail before writing the prompt.
Composition-first anchoring — Defining the visual structure (for example an editorial still life with diagonal placement and controlled directional lighting) before describing the product, so the aesthetic does not become unstable. Apply: Open the prompt with the composition and only then describe the product.
Exact product identity statement — An explicit instruction that the product must match the provided references exactly, covering proportions, silhouette, stitching, panel geometry, sole curvature and overall construction. Apply: Add a second prompt block listing those attributes as must-match. The video says this prevents redesign drift and reinterpretation.
Dedicated typography block — A separate, high-priority prompt section that lists the exact words that must appear, where they appear, their alignment and orientation, and that there be no misspellings, extra letters or warped glyphs. Apply: Write the exact words, their placement and their alignment as their own block, not inside the general product description.
Specific color naming (true color accuracy) — Naming each component's color precisely instead of using generic terms, for example deep cobalt blue upper, neon lime laces, off-white mid-sole and mint heel lining. Apply: Replace general color words like 'blue shoe' with component-level color descriptions to reduce reinterpretation and keep brand identity intact.
Lighting and harmony definition — Describing light direction and tonal balance so the environment supports the product instead of competing with it. Apply: Include a lighting block that states direction and tonal balance after the color block.
Optical control — Specifying camera and focus qualities such as shallow depth of field, tack-sharp focus on the shoe and its typography, refined tonal balance and no oversaturation. Apply: Add these optical instructions so the result feels premium and controlled, not over-processed.
Negative constraints — Explicit prohibitions (no extra logos, no duplicated text, no invented branding, no distorted lettering) that block the model's tendency to improvise. Apply: End the prompt with these exclusions to reduce correction cycles and protect brand integrity.
Kling 3.0 Omni multi-reference reinforcement — Using Kling 3.0 Omni with several image references that stay active throughout generation, preserving proportions, materials and typography across frames. Apply: In the video generator, skip the start image and upload the same clean white studio references used for the stills.
Minimal structured video prompt — A short video prompt that describes the scene and camera movement and keeps product identity consistent, without redefining typography or altering branding. Apply: Describe the action, such as a model picking up the shoe and bringing it toward the camera, and let the references hold the typography.
The video says composition goes first because, if it is not anchored, the model prioritizes the object itself and the overall aesthetic can become unstable. In its account, ordering controls what the model treats as primary.
The video says reference quality limits typography accuracy: if the source images are weak, accuracy becomes much harder to control. It asks for multiple angles, clean lighting and sharp detail.
The same clean white studio references used to generate the stills are reused in the video generator in place of a start image, so one reference set links the image and video stages.
In the video stage the prompt stays minimal (scene, camera movement, consistent identity). It deliberately does not redefine typography or alter branding, because the references hold those anchors.
The video swaps vague color words for specific ones (deep cobalt blue upper, neon lime laces, off-white mid-sole, mint heel lining). It says this reduces unwanted reinterpretation and keeps brand identity intact.
The video states typography-specific negatives (no misspellings, no extra letters, no warped glyphs) and general negatives (no extra logos, no duplicated text, no invented branding, no distorted lettering). The first set belongs in the typography block and the second in the closing constraints.
The test of two very different environments (sandstone vs. terracotta, different materials and color balance) with the same formula is presented as showing the structure is not tied to one scene.
«The goal is not to create a complicated prompt. The goal is to create a structured prompt.»
— 01:11
«Typography should not be casually mentioned inside the general product description. It needs to be treated as a priority anchor.»
— 02:19
«Models tend to improvise, so you need to block that behavior directly.»
— 03:20
«Composition first, exact identity, typography control, true color, lighting harmony, optical clarity, negative constraints.»
— 03:33
«It's not about writing more, it's about writing in the right order.»
— 03:41
«It's not the length of the prompt, it's the order and priority of information.»
— 04:35
«Perfect typography is not an accident. It's a result of structure and reinforcement.»
— 05:50
Reception
Viewers overwhelmingly praise the video as a genuinely useful, non-promotional tutorial that solved long-standing problems, with the remaining comments being constructive follow-up questions, feature requests (Spanish dubbing, product labels, character consistency), and brand replies.
A compact, actionable walkthrough built around an ordered seven-part prompt formula, shown on a single running-shoe example. Its distinctive move is reusing one set of studio references to carry typography consistency from stills into Kling 3.0 Omni video.

06:08