A complete workflow for turning protein and nutrient data into clear animated food comparisons using AI video, templates, and tight motion rules.
Most nutrition comparisons fail for a boring reason: the numbers are correct but invisible. A table showing that cooked chicken breast carries roughly three times the protein of cooked lentils per 100 grams is accurate, useful, and emotionally flat. Nobody stops scrolling for it.
Motion changes the arithmetic of attention. A bar that grows, a plate that splits, a ring that fills — these give magnitude a shape the eye can read in about two seconds. That is the actual job of an animated infographic: not decoration, but instant comprehension.
This guide covers the full production of a protein-and-nutrient comparison video: cleaning the data, choosing a storyboard that fits the claim, writing prompts that produce usable frames, keeping visual consistency across a series, and exporting platform-specific cuts. It is written for creators, dietitians, educators, and brand teams who want a repeatable process instead of a one-off animation.
Why motion beats a static nutrient table
Static charts ask viewers to do work: locate the axis, compare heights, convert units, remember what the previous chart said. Motion performs the comparison for them. When two bars grow in the same frame, the gap becomes the message, and the size of that gap is what people remember.
Three advantages are worth designing around.
Ranking becomes legible in real time. A sorted bar race reorders itself as values change, so viewers see not only who wins but by how far. Relational information survives much longer in memory than an isolated number.
Several dimensions fit in one frame. Protein, fat, carbohydrate, and fiber can live in separate channels — bar height, ring fill, color saturation — without turning into a spreadsheet. One clip can carry four metrics that would need four separate charts in print.
Context can be layered. A serving-size plate, a source label, and a per-100-gram toggle can all sit inside the same composition. This is exactly where static infographics run out of space and start using six-point type.
One caution worth repeating: animation implies change over time. A growing bar should represent a comparison between foods, not a claim that a food's protein content is rising while you watch. Label the axis and the basis clearly enough that nobody reads a measurement as a process.
Build the data sheet before any design
Every visual decision downstream inherits the quality of your table. Build it in a spreadsheet, in plain text, and only then open a design or video tool. Six columns are enough: food name, state (raw or cooked), protein per 100 g, fat per 100 g, carbohydrate per 100 g, plus a realistic serving size and its protein value.
Compare on two baselines
Per-100-gram values are the fair scientific basis. Per-serving values are what people actually eat. Both matter, and they sometimes tell opposite stories.
Per 100 grams, almonds look like a protein powerhouse at roughly 21 grams. Per a typical serving of about 28 grams, that becomes roughly 6 grams. A 150-gram portion of cooked chicken breast delivers roughly 45 grams. Neither number is wrong — the baseline changes the conclusion.
Show both where space allows, or choose one and state it in large type for the entire clip.
Separate cooked from raw
This single mistake inflates more comparisons than any other. Cooked lentils and raw lentils differ by roughly a factor of three once water is absorbed. Rice, pasta, beans, and most grains behave the same way. Record the state in the sheet and repeat it on screen; a note reading cooked, per 100 g costs four seconds of design work and prevents an accusation of exaggeration.
Round late, round once
Keep two decimals in the sheet and round to whole grams for on-screen numerals. Precision beyond that is noise that slows reading, and it invites arguments about the third digit instead of the story.
Cap the cast
Eight to twelve foods is the sweet spot for a 45 to 60 second comparison. Beyond twelve, a bar race becomes a blur and viewers stop tracking individual items. If your topic genuinely needs thirty foods, make three videos by category rather than one crowded chart.
A workable starter set for a protein comparison, cooked where applicable and approximate: chicken breast (~31 g per 100 g), salmon (~20 g), firm tofu (~16 g), eggs (~13 g), cottage cheese (~11 g), Greek yogurt (~10 g), lentils (~9 g), chickpeas (~9 g), almonds (~21 g), quinoa (~4 g). Verify every value against a current food composition database for your region before publishing; cuts, brands, and formulations vary.
Choose a storyboard format that fits the claim
You do not need a novel structure. Four formats carry almost any nutrient comparison, and each has different production demands.
The sorted bar race
Bars enter at zero, grow in a staggered sequence, and reorder into a final ranking. Strongest for protein, fiber, calories, and sodium, where a single number drives the story.
Production note: generate bars as flat rectangles on a neutral background with no text baked into the frames. Add numerals, units, and labels in your editor so they stay crisp and editable when a value changes.
The split plate reveal
A top-down plate is divided; one side swaps foods while macro rings around the frame fill to the corresponding values. This handles the same-calories-different-composition argument better than a bar chart because the plate keeps portion size visible.
Production note: lock the camera to a strict top-down position and keep lighting identical across swaps. Any drift in shadow direction reads as an error, and viewers notice errors before they notice numbers.
The food-to-macro transformation
The whole food dissolves into a stack of macro blocks — protein cubes, fat droplets, carbohydrate chains. This is the most stylized and most shareable option, and it needs the most restraint. Transformation sequences can imply that a food converts into something else nutritionally, which is misleading if unlabeled. Use it as a transition into a numbers card, never as the argument itself.
The numbers-first card
Sometimes the cleanest format is a plain full-frame number, held for two seconds, with the food name and the basis beneath it. Three of these in sequence, then the comparison. Cheap to produce, easy to caption, and unbeatable for accessibility.
Lock a visual code across every scene
Consistency is the difference between a video and a slideshow of unrelated clips. Four rules cover most of it.
One color per macro, permanently. Protein teal, carbohydrate amber, fat rose, fiber muted green. Reuse identical hex values in titles, bars, and captions. Viewers learn the code in the first ten seconds, then read faster for the rest of the clip.
One camera grammar. If scene one is locked top-down, scene four cannot be a handheld three-quarter angle. Write the framing into every prompt rather than hoping the model remembers.
One type system. A single family, two weights, three sizes. Numbers large, labels small, source line smaller. Never introduce a display font mid-video for emphasis.
One surface. Same background color and texture throughout. Changing surfaces between foods makes a single comparison feel like separate shoots.
A practical trick: approve a three-frame lookbook before animating anything — one title frame, one comparison frame, one numbers frame. Treat those as reference for every generated clip. Starting from an existing video template and adapting it is faster than inventing a look from zero.
Write prompts for clean plates and graphic layers
Generative video is excellent at texture, surfaces, and camera motion, and noticeably weaker at precise typography. Design around that constraint instead of fighting it.
The prompt stack
A reliable prompt covers six things, in order:
- Subject and state — what the food is, cooked or raw, sliced or whole.
- Framing — top-down locked, macro 100 mm, wide tabletop.
- Lighting and surface — soft studio key, matte charcoal or light stone.
- Motion — push in, static lock, or slow orbit; be explicit if you want no movement at all.
- Graphic elements — matte bars, translucent rings, empty label zones.
- Exclusions — no text, no logos, no hands, no extra props.
That last line matters more than beginners expect. Without it you get invented packaging, floating numbers, and stray cutlery competing with your data.
Example prompt: comparison plate row
Top-down locked studio shot on a matte charcoal surface, ten small white ceramic dishes arranged in a single row, each holding one cooked ingredient — sliced chicken breast, grilled salmon, firm tofu cubes, halved boiled egg, Greek yogurt, cottage cheese, cooked lentils, cooked chickpeas, almonds, cooked quinoa. Soft diffused key light from the left, subtle contact shadows, no camera movement, shallow depth of field at the dish edges. Clean empty space above each dish for overlays. No text, no labels, no hands, no packaging, no props.
Example prompt: ring background
Abstract dark slate background with three thin translucent rings of different radii, centered slightly off-frame, matte finish, teal, amber, and rose at low opacity, very slow continuous rotation, no camera shake, even studio lighting, subtle grain. No text, no numbers, no objects.
Generate the food plates and the graphic layer separately. Compositing them yourself keeps numerals sharp and lets you revise a value without re-rendering an entire scene. Locking still assets first with the AI image generator is usually faster than iterating on video and hoping the frame settles.
Keep a prompt file per project with the text that worked. Reusing proven language is the shortest route to consistent output, and a shared prompt library makes that reuse systematic rather than remembered.
Motion, captions, and sound rules
Bad motion in data video almost always means too much, too fast, or too linear.
- Reading time beats runtime. Give each on-screen value at least 1.2 seconds of stillness before the next change. Two numbers per second is unreadable.
- Ease out, do not snap. Bars should decelerate into their final value. Linear growth reads as a loading bar, not a measurement.
- Animate four things maximum per scene. Beyond that, attention fragments.
- Hold the final ranking. Two full seconds on the finished comparison, longer if there is a takeaway caption.
- Match counters to shapes. If a bar grows to 31 grams, the numeral must arrive at 31 on the same frame.
- No decorative loops during reading. Slow orbits and drifting particles belong in transitions, and they are hostile over text.
Assume the video will be watched muted, because it usually will be. Burn in captions for every spoken claim, keep them inside the safe area, and maintain strong contrast against the background. On a 1080x1920 vertical export, headline numbers should be at least 90 pixels tall to stay legible on a phone at arm's length.
On audio, resist scoring the whole clip. A low bed with a soft tick on each bar arrival is enough. Voiceover should sit clearly above the music; if viewers strain to hear, the data is lost. Accessibility also means describing the comparison in text — a caption stating the basis (grams of protein per 100 g, cooked) and the source does more for comprehension than any visual flourish.
A production workflow that scales into a series
1. Write the one-sentence claim. For example: per 100 grams, firm tofu delivers about half the protein of chicken breast, but with fiber and no cholesterol. Everything in the video serves that sentence, and the video ends when it is proven.
2. Build the data sheet. Foods, state, per-100-gram values, serving values, source name, date checked. Two decimals, one source per column.
3. Design the lookbook. Three frames. Approve colors, type, surface, camera angle.
4. Generate still assets. Plates, backgrounds, graphic layers. Lock the ones you will reuse.
5. Assemble a rough animatic. Fifteen seconds of flat shapes carrying the real numbers, no rendering. If the sequence does not communicate here, better prompts will not save it.
6. Render scenes. Short clips, three to five seconds each, produced with the AI video generator and prompted from your lookbook language rather than from scratch.
7. Composite and quality-check. Add numerals, labels, units, and the source line in the editor. Then check every visible number twice: once against the sheet, once against the on-screen value. This step catches more errors than any other.
Then export three cuts from the same timeline. Vertical 9:16 at 20 to 30 seconds for short-form, built around the single strongest gap. Horizontal 16:9 at 60 to 90 seconds for explainers, adding the per-serving view and one counterintuitive nuance such as fiber or saturated fat. A six-second silent loop showing only the final ranking for landing pages, waiting-room screens, or clinic displays. Keep a silent version of each for presentations.
Mistakes that quietly break a nutrition video
Mixing cooked and raw weights. A raw-versus-cooked comparison can triple an apparent difference without anyone lying outright. State the state.
Using per-100-gram values for foods eaten in small portions. Nuts, seeds, and hard cheese look very different once realistic servings appear. Add the serving view.
Baking text into generated frames. Regenerating a scene to fix a typo wastes hours. Keep numbers in the edit.
Too many foods, too tiny. Twelve is generous. Eight is comfortable. Six is often better.
Decorative footage that contradicts the data. A sizzling steak shot under a slide about plant protein creates a mixed message the numbers cannot fix.
Loose health claims. High protein needs a definition — a common threshold is roughly 20 percent of calories from protein, or about 10 grams per 100 grams. If you use the phrase, define it on screen.
No source line. Even a short label naming the database and the date checked raises trust significantly and costs four seconds of design work.
Ignoring the thumbnail. At feed size, silhouettes matter more than detail. If two foods render identically, differentiate with dish shape, garnish, or a subtle color shift rather than a label you will have to re-render.
Frequently asked questions
How long should a nutrient comparison video be?
Twenty to sixty seconds for social, ninety seconds maximum for a full explainer. If you need longer, split by category — protein sources, then fats, then fiber — instead of extending one crowded chart.
Can AI video tools render accurate on-screen numbers?
Reliably, no. Treat generated footage as plates and backgrounds, then add numerals and labels in an editor. This also keeps values editable when your data changes.
Should I compare per 100 grams or per serving?
Both if you can. Per 100 grams is the fair scientific basis; per serving is what viewers actually eat. If you must choose one, use per serving for consumer content and label it conspicuously for the entire clip.
How many foods should appear in one clip?
Eight to twelve. A sorted bar race with more than twelve entries produces visual noise and viewers lose track of individual items.
What if two foods look identical in my render?
Differentiate with dish shape, garnish, or a small color shift, not with labels baked into the frame. Distinct silhouettes matter more than fine detail at thumbnail size.
Do I need a voiceover?
No. A well-paced comparison with captions and restrained sound design often performs better muted. If you add narration, keep each sentence to one claim and one number.
How do I keep a series consistent across episodes?
Save your lookbook, hex values, type sizes, prompt text, and export presets as a project kit. Every new episode starts from that kit rather than a blank canvas.
Can I use this workflow for non-food data?
Yes. Anything with a small set of comparable values works: exercise minutes, sleep hours, water use, screen time, study hours. The constraints that matter are the same — a limited cast, one clear claim, and numbers added in the edit.
Build your first comparison this week
Animated nutrition comparisons are not primarily a technical problem. They are a discipline problem: pick one claim, verify eight to twelve values against a real database, approve three frames, then animate only the motion that makes magnitude visible. The rest is editing, and editing is where the video becomes trustworthy.
Start with the Orelon AI video generator to render plates and motion backgrounds, lock consistency with the template library, and keep every prompt that worked in your own notes. One well-built comparison usually becomes a series, and a series built on the same visual code becomes something viewers recognize before they read a single number. If you want to see how other creators structure their sequences, the Orelon blog is a good next stop.

