Food Memory
Why Repeated Meals Should Get Easier to Log: How Food Memory Works
A familiar bowl of oats should not feel completely new every morning. Food Memory keeps inspectable portion hints for normalized ingredients—without pretending that a stored value proves today's amount.
What is Food Memory?
Food Memory is Food Focus's optional portion-recall feature, enabled by default on each device. Its Settings toggle is stored locally rather than synced account-wide. In the normal single-meal flow, saving a meal can store the current grams for each ingredient the analysis recognizes and normalizes. If a later photo, text, or voice log recognizes the same ingredient but does not make the current amount clear, that stored portion may be supplied as a personal hint.
It is intentionally narrower than a prediction system. Food Memory does not calculate an “average serving,” build a frequency-weighted profile, or guarantee what is on today's plate. It keeps one inspectable current value for a matched ingredient. You can rename an account entry, change its grams, delete it, clear the full list, or turn future memory use off on the current device.
What exactly does Food Memory save?
In the standard in-app logging flow, Food Memory updates after the analyzed meal is reviewed and saved. Its unit is a recognized, normalized ingredient and a gram value—not the photograph itself and not a general label such as “my usual breakfast.” Siri logging is an exception: it can analyze, save, and update memory in the background without showing the normal pre-save review screen, so review that logged meal afterward.
| Food Memory does | Food Memory does not |
|---|---|
| Keep one current gram value for a recognized, normalized ingredient | Keep a running average of all portions |
| Update after a save when Food Memory is enabled on that device | Update merely because an analysis was started and abandoned |
| Keep up to 200 recently updated entries for the account | Keep an unlimited archive of every historical portion |
| Make the current name and grams visible and editable | Hide the value inside an unexplained personal score |
| Allow individual deletion or clearing the full list | Guarantee recognition or matching for every food name |
“Recognized and normalized” matters. The analysis must identify an ingredient and connect it to a consistent ingredient identity before a later log can match it. Similar names do not guarantee a match, and a mixed dish may not expose every ingredient separately.
Does Food Memory learn your average portion?
No. In the normal single-meal flow, Food Memory uses a replacement rule: the newly saved portion for the same recognized, normalized ingredient replaces the value stored before it. Earlier values are not combined into an average, and a portion is not weighted more heavily because you logged it more often. If you manually edit an entry, that edited gram value becomes the current hint.
- 1
Save a reviewed meal
The saved analysis contains ingredients and their logged portions.
- 2
Match the normalized ingredient
If that ingredient already has a Food Memory entry, the app identifies the existing entry.
- 3
Replace the stored portion
The portion from the new single-meal save becomes the current value. The previous value is not averaged into it.
This makes the usual flow easy to interpret: the value you see is the current saved or manually edited value that may be offered as future context. It should not be described as your typical, preferred, recommended, or statistically predicted serving.
When can a remembered portion affect a new food log?
A remembered portion may be used when a future photo, text, or voice analysis recognizes the ingredient but the current input leaves its amount ambiguous. The stored value is extra context for the estimate. It is not stronger evidence than a clear current amount supplied by you.
| Current log | Role of Food Memory | Why |
|---|---|---|
| A photo includes oats but the amount is unclear | May supply the latest saved oats portion as a hint | The current image does not establish a clear amount |
| Text says “oats with yogurt” without quantities | May supply remembered portions for matched ingredients | The foods are named but the amounts are ambiguous |
| Voice says “80 grams of oats” | The stated current amount should be reviewed as current evidence | The description already provides a clear amount |
| The new meal contains a different or unrecognized ingredient | No matching remembered portion is guaranteed | Food Memory depends on ingredient recognition and normalization |
Portion uncertainty is not unique to Food Memory. Reviews of image-based dietary assessment describe portion or volume estimation as a difficult stage, especially when a single image lacks depth or ingredient detail. A remembered value can provide personal context, but it cannot turn that image into a direct measurement. See the scoping review in the Journal of Medical Internet Research (opens in a new tab).
Example: how the latest portion changes
The following is an illustration of the documented replacement rule, not a prediction or validation result.
- 1
Monday: save 80 g of rolled oats
If rolled oats are recognized and normalized, Food Memory can keep 80 g as the current latest portion.
- 2
Thursday: save 60 g of rolled oats
The 60 g portion replaces 80 g. Food Memory does not store 70 g as an average.
- 3
Saturday: log “oats” without an amount
If the ingredient matches and the amount remains unclear, 60 g may be supplied as a hint. You still check whether Saturday's bowl was actually 60 g.
How can you manage what Food Memory remembers?
Food Memory is designed to be inspectable. You can open the list, see each normalized ingredient and gram value, edit the ingredient name or grams, delete one entry, or clear everything. The store keeps at most 200 recently updated entries per account.
- Review the ingredient identity. Confirm that the entry represents the food you expect.
- Edit the name or grams. A valid manual edit becomes the current value available to later analyses.
- Delete stale context. Remove an individual entry if the amount no longer provides a useful hint.
- Clear all entries. Remove the full account list when you want to rebuild it from later meal saves.
- Turn future use off on this device. The locally stored Settings toggle stops use and updates from that device without deleting account entries or changing another device's setting.
- Correct future meal estimates. Even a useful remembered entry does not remove the need to check the current meal.
For the product-level description of these controls, see the Food Memory feature page. The sources and methodology page explains how remembered portions fit alongside AI-assisted nutrition estimates.
What does Food Memory not know?
A latest portion is useful only to the extent that today's ingredient and serving resemble the saved one. Food Memory does not observe why a portion changed, whether a recipe was reformulated, how much was left uneaten, or whether the current serving uses the same preparation.
- It does not guarantee recognition. An ingredient must be recognized and normalized before it can be stored or matched.
- It does not remember an average. The normal single-meal flow replaces the current value rather than averaging earlier portions.
- It does not measure today's portion. The stored amount is a hint when the current amount is unclear.
- It is capped. The account keeps the 200 most recently updated entries rather than an unlimited history.
- Batch saves need review. Pasting several meals together can sum duplicate normalized ingredients into one stored value.
- Siri skips the normal review screen. A Siri meal can be saved and update memory in the background, so check the resulting log afterward.
- It does not reveal hidden ingredients. Oil, sauces, fillings, and recipe changes can still alter the meal estimate.
- It does not make nutrition guidance medical. The resulting calorie and nutrient values remain informational estimates.
Treat Food Memory as a shortcut for repeated context, not as proof. Review the current foods and quantities before saving, especially when your bowl, recipe, appetite, goal, or serving has changed.
How can you try Food Memory responsibly?
In the standard app flow, begin with a meal whose ingredients and amounts you can review. Correct the analysis before saving it, because the saved ingredient portion can become the current context. On a later ambiguous photo, text, or voice log, check whether any remembered value fits the current serving and edit the result when it does not. If you log through Siri or paste several meals, inspect the saved meal and Food Memory afterward.
Build useful context
Save a reviewed portion. Keep the next estimate editable.
Food Focus includes the first 10 meal logs before Pro is required. Use familiar ingredients, inspect Food Memory, and decide whether latest-portion context helps your routine.
Questions about Food Memory and remembered portions
What exactly does Food Memory remember?
When Food Memory is enabled, the normal single-meal save flow stores the current grams for ingredients Food Focus recognizes and normalizes. Existing entries are replaced rather than averaged, and a manually edited name or gram value becomes the current hint.
Does Food Memory learn my average portion?
No. In the normal single-meal flow, newly saved grams replace the prior value for that recognized, normalized ingredient. A manual edit can set the current hint, while a multi-meal paste can sum duplicate normalized ingredients. It is not an average, trend, or frequency-weighted prediction.
When can Food Memory affect a new meal estimate?
A later photo, voice, or text analysis may receive the remembered portion as context when it recognizes the ingredient but the current input does not provide a clear amount. A clear current amount should take priority.
Can I edit or delete remembered portions?
Yes. You can view entries, rename a normalized ingredient, change its grams, delete one entry, or clear the full list. The account store is capped at 200 recently updated entries.
Can I turn Food Memory off?
Yes. Food Memory is enabled by default on each device. Its locally stored Settings toggle stops remembered portions from being used and updated from that device; it does not delete account entries or change the setting on another device.
Does a remembered portion make the next estimate accurate?
No. It adds personal context, but today's serving may be different and the ingredient still has to be recognized. Review and correct the current meal before relying on it.
Sources and further reading
These sources support the general explanation above. They do not endorse Food Focus, and research methods do not automatically establish the accuracy of any specific consumer app.
- Food Memory: A Calorie Tracker That Remembers (opens in a new tab)Food Focus
The canonical product description of what Food Memory stores, when it can supply context, its replacement rule, and user controls.
- Food Focus nutrition sources, methodology and limitations (opens in a new tab)Food Focus
The product-specific methodology for Food Memory, AI estimates, corrections, and the distinction between context and measurement.
- AI Food Scanner: Photo, Voice and Barcode (opens in a new tab)Food Focus
The five-input Food Focus logging workflow in which remembered portions can support ambiguous photo, voice, or text estimates.
- Advancements in Using AI for Dietary Assessment Based on Food Images: Scoping Review (opens in a new tab)Journal of Medical Internet Research
A review explaining why single images lack exact portion, ingredient, and preparation information and why additional context can matter.