In this article (5 sections)
The camera is an input device, not a guarantee that two tools use the same information. Before comparing speed or free access, identify what the tool is reading and where the resulting nutrition comes from.
What does each scanner actually do?
| Input | Information supplied | Useful task | What still needs checking |
|---|---|---|---|
| Meal photo | Appearance of food in an image | Starting an ingredient breakdown for a plate | Hidden ingredients, cooking state, grams and food matches |
| Barcode | Product identifier | Finding an existing packaged-food record | Exact variant, current formulation, serving basis and amount eaten |
| Nutrition-label photo | Printed values and units | Entering a product missing from search | Digits, serving versus per-100 g basis, units and absent fields |
| Manual entry | Values you supply | Preserving a known label or recipe calculation | Arithmetic, quantity and source note |
These are descriptions of input methods. They are not a claim that every app offers all four tools for free or on every platform.
When is a meal photo useful?
A photo can help organize a visible plate into candidate ingredients. It cannot weigh food or reveal the full recipe inside a sauce. Supply known quantities and additions, such as cooking oil, instead of asking the image to prove them.
For example, a photo might suggest rice and chicken while missing oil stirred into vegetables. The correction is to use the known rice weight, confirm the chicken's preparation state and add the actual oil amount. Our photo logging guide separates visible evidence from assumptions; AI correction examples show how a plausible first result can change.
Cronometer's official photo logging instructions also ask users to review matched foods. That is useful context for the method, not independent evidence of a particular accuracy rate.
Why can barcode results disagree with the package?
A barcode lookup can return a record for a different market, older formulation or incorrect serving entry. Finding a record establishes an identifier match in that database, not that every value is current. Compare the actual package before accepting the result.
If an invented label says 200 kcal per 40 g but a result shows 200 kcal per 100 g, the difference is the serving basis. At 60 g those entries produce 300 kcal and 120 kcal respectively. Repeatedly rescanning the same barcode will not fix that mismatch. Use the label's correct basis or a manual record.
Calorie.one uses Open Food Facts for packaged-product lookup alongside its other sources; the methodology explains those boundaries. A packaged-food record and a USDA generic ingredient record should not be described as the same source.
What can go wrong when reading a nutrition label?
Glare, a curved package or small print can obscure a decimal point or make two columns merge. Check whether the values describe one serving, 100 g, the whole package or the prepared food. Read the serving amount as well as the calorie number. The FDA label guide explains these reference quantities.
Review missing fields too. If the package does not state a nutrient, its absence is not a measured zero. A readable calorie value does not authorize inventing protein, fiber or sodium. See reading US labels for portion arithmetic.
When should you choose manual entry?
Use manual values when you already have a reliable package or a completed recipe calculation, or when recognition keeps selecting the wrong entry. Check the units once, enter your actual amount and confirm the preview. No new estimate is needed merely because the app offers AI.
Calorie.one supports food search, barcode lookup, label-photo review, text/photo analysis and manual nutrition entry in the relevant logging flows. Access to AI follows the account allowance, and availability can differ by device. Start with the web app; an account is required to save a meal. For a known recipe, use the calculation method before choosing a camera tool.