Batch-cooking desserts: creams, flans and panna cotta without gelatin

Two numbers will stop you ruining these desserts. Agar gels from roughly 0.2 % of the weight of the liquid, and it has to be boiled to work at all: below 85 °C it will not even dissolve. Cornflour is a different animal. It thickens without gelling, at around 4 % of the liquid for a cream that coats the back of a spoon. The two ingredients do different jobs, and confusing them explains most failed textures.

Three things this article will not sell you. A homemade dessert pot is not automatically less sugary than a shop-bought one: the flan weighed out below comes to 12.4 g of sugars per 100 g against 16.0 g for a chilled egg flan, a real gap but a modest one. A coconut-cream panna cotta delivers 11.2 g of saturated fat per pot, more than half the NHS daily reference for an adult woman. And these desserts do not keep for a vague “several days”: the Food Standards Agency sets 48 hours for chilled leftovers, and the food has to go into the fridge within two hours.

What agar does, and what gelatin used to do

Gelatin is hydrolysed animal collagen. It melts around body temperature, which is exactly why it feels like it dissolves in the mouth. No plant gelling agent reproduces that sensation, and accepting this is the first step before you substitute anything.

Agar is a polysaccharide extracted from red seaweed, authorised as a food additive under the code E 406. The re-evaluation opinion published by EFSA’s scientific panel in 2016 describes its physical behaviour precisely. It is insoluble in cold water and only dissolves above its gel melting point, at 85 °C or higher. Once dissolved, the sol can be cooled well below that temperature before it sets, and it then needs strong reheating to melt again. This gap between setting temperature and melting temperature has a name, gelation hysteresis, and one very practical consequence: an agar panna cotta does not collapse in a warm room in summer, unlike a gelatin one.

The gelation threshold sits at around 0.2 % according to the same opinion, which is very low indeed. In practice a level teaspoon weighs about 2 g; in a litre of liquid that gives 0.2 %, just on the threshold. This is why vague recipes calling for “two teaspoons of agar” produce erratic results: depending on whether the spoon is level or heaped you swing between 0.2 % and 0.4 %, and at 0.4 % you get a brittle gel you can slice with a knife rather than a creamy dessert.

Two chemistry details matter as well. Agar undergoes acid hydrolysis as soon as it is heated for any length of time below pH 5.5, which destroys its gelling power. Adding lemon juice or an acidic fruit purée to the boiling liquid therefore sabotages the set. Add the acid after boiling, or layer the coulis separately. Tannic acid, present in fruits such as apple and prune, also blocks the gelling process outright.

EFSA further notes that agar is very poorly absorbed and only slightly fermented by gut microbiota. The highest exposure estimate in the opinion reaches 26 mg per kilogram of body weight per day in toddlers at the 95th percentile, far below the doses tested in toxicology. No numerical acceptable daily intake was considered necessary: the Scientific Committee for Food allocated agar an ADI “not specified” back in 1989. At home you are working in grams per litre, so nowhere near those levels. Worth noting all the same: daily doses of 4 to 16.5 g have been used as a mild laxative, which tells you what to expect from a persistent excess.

Cornflour is not a gelling agent

This is the most common confusion. Cornflour and potato starch thicken: their granules swell in heat, absorb water and raise viscosity. They do not build a network that stands up outside a mould. A cornflour cream stays a cream; it will not turn out.

In practice, count 40 to 45 g of cornflour per litre of liquid for a dessert cream that coats the spoon, so roughly 4 %. Below 30 g per litre you get a sauce. Above 60 g the texture turns pasty and floury in the mouth. The starch has to be slaked in a little cold liquid before it goes into the hot pan, otherwise it forms lumps that will never dissolve.

One technical point explains creams that weep water after three days: starch retrogrades. Its chains slowly reorganise in the cold and push water out, a phenomenon called syneresis. A study published in Polymers in 2025 measured this on milk puddings containing 0.3 % hydrocolloid and 1 to 5 % modified waxy corn starch. Syneresis increased in every pudding over two weeks of refrigerated storage, without exception. The authors also found that lower starch concentrations, around 1 % rather than 5 %, reduced syneresis at 4 °C after 7 and 14 days, and that increasing the starch actually lowered gel strength. Loading up on cornflour to get a firmer texture is therefore counterproductive on both counts.

The same study compared five agents. Agar and kappa-carrageenan produced the most rigid and strongest gels, while gellan gum and iota-carrageenan were better at preventing syneresis. Sucrose significantly increased gel strength in puddings containing kappa-carrageenan, gellan gum and agar. Cutting the sugar in a gelled recipe therefore changes the texture too, not just the taste.

The other plant gelling agents, and what they are worth

Pectin belongs to the jam world and does not behave like agar. It comes in two families. High-methoxyl pectin needs a heavily sugared, acidic environment to set: the confection gels studied in 2013 with HM pectin and grape juice had a soluble solids content of 73 % and pH values between 2.5 and 3.3. That is not the composition of a dessert cream. Low-methoxyl pectin gels in the presence of calcium ions rather than sugar, which makes it usable in lightly sweetened preparations. This is the principle behind reduced-sugar jams. For a flan or a panna cotta, agar remains far simpler to dose.

EFSA re-evaluated pectins E 440i and E 440ii in 2017 and maintained an ADI “not specified”. Pectins are not absorbed intact but are extensively fermented by gut microbiota, and no adverse effects were reported in a chronic toxicity study in rats up to 5,000 mg per kilogram per day, the highest dose tested. A follow-up opinion published in 2021 tightened the specifications, in particular for arsenic, lead, cadmium and mercury.

Carrageenan, E 407, deserves a mention because it turns up in a great many shop-bought plant drinks and industrial desserts. Its regulatory position is less comfortable than agar’s. In its 2018 opinion EFSA concluded that the group ADI of 75 mg per kilogram of body weight per day for carrageenan and processed Eucheuma seaweed should be considered temporary, given uncertainties around chemistry, exposure and toxicological data, and asked for the database to be improved within five years of publication. The opinion reports a refined exposure of up to 758.6 mg per kilogram per day in toddlers under a brand-loyal scenario. None of this makes carrageenan a poison, but it does justify reading the ingredient list on the plant drink you use as a base, especially since that base is 80 to 90 % of the weight of the finished dessert.

A less familiar route: blending agar with locust bean gum reduces the loss of liquid from the gel during handling, transport and storage, as the EFSA agar opinion notes. Work published in 2024 on agar/locust bean/xanthan hydrogels measured the degree of syneresis dropping from 2.33 % to 0.52 %, with hardness falling from 702 to 343 g and springiness rising to 1.00. That is exactly the direction you want when replacing gelatin: less hard, more elastic, less water separating out.

Three weighed recipes, with their calculated composition

The compositions below are calculated ingredient by ingredient from the ANSES Ciqual 2020 food composition table. They are not laboratory measurements of the finished product, but they are exact as a sum of the ingredients, which is plenty for knowing where you stand.

Chocolate cornflour cream, 8 pots

For 1 litre of calcium-fortified soya drink, 80 g of 70 % dark chocolate, 45 g of cornflour, 30 g of sugar and a pinch of salt. Yield: about 1,155 g, so eight 145 g pots.

Slake the cornflour in 100 ml of cold drink taken from the litre. Heat the rest with the broken chocolate and the sugar, stirring until the chocolate has melted. Pour in the slaked cornflour while whisking, then hold at a gentle boil for three to four minutes, stirring constantly. That time matters: cornflour that has not boiled keeps a raw taste and has not given up all its viscosity. Add the salt off the heat, fill the pots, let them cool then refrigerate.

Calculated composition per 100 g: 104 kcal, 3.7 g protein, 9.8 g carbohydrate of which 5.7 g sugars, 5.0 g fat of which 2.2 g saturates, 1.8 g fibre, 89 mg calcium. Per 145 g pot: 150 kcal, 8.3 g total sugars and 3.2 g saturates. Free sugars, meaning the added sugar plus the sugar in the chocolate, come to 5.5 g per pot.

For comparison, the chilled chocolate dessert cream in the Ciqual table shows 16.5 g of sugars per 100 g. The homemade version here is three times lower, but that is because it contains 30 g of sugar per litre, not because it is homemade. Double the sugar, as plenty of recipes do, and the gap vanishes.

Coconut and soya panna cotta with agar, 6 glasses

For 400 g of coconut cream, 400 g of plain soya drink, 60 g of sugar, 3.5 g of agar and a split vanilla pod. Yield: about 860 g, so six 145 g glasses. The 3.5 g of agar in 800 g of liquid works out at 0.44 %, deliberately above the threshold because the fat in the coconut interferes with network formation.

Combine everything cold in the pan, agar included. Bring to a proper boil and hold it simmering for one to two minutes while whisking: this is where the agar dissolves, and nothing else compensates for a missed boil. Remove the pod, pour into the glasses straight away, because setting begins as soon as the temperature drops towards 40 °C. Let them cool, then refrigerate for at least three hours.

Calculated composition per 100 g: 133 kcal, 2.4 g protein, 8.9 g carbohydrate of which 8.1 g sugars, 9.5 g fat of which 7.8 g saturates. Per 145 g glass: 190 kcal and 11.2 g of saturated fat.

That last figure is the real story of this recipe. The NHS advises no more than 30 g of saturated fat a day for men and 20 g for women, and explicitly lists coconut oil and coconut cream among the plant sources of saturates. One glass therefore covers 56 % of the daily reference for an adult woman. That is comparable to the shop version, which shows 8.89 g per 100 g in the Ciqual table against 7.8 g here. A coconut panna cotta is not a light dessert; it is a rich dessert whose richness simply comes from somewhere else.

If you want to bring that number down, swap 200 g of coconut cream for 200 g of soya drink and raise the agar to 4 g: saturates fall to around 4 g per glass, the texture becomes firmer and cleaner, and the coconut flavour largely disappears.

Baked egg flan, 6 ramekins

For 700 g of semi-skimmed milk, 200 g of whole egg, meaning four medium eggs, 90 g of sugar and a vanilla pod. Yield: about 990 g, so six 165 g ramekins. There is no added gelling agent here: the egg proteins coagulate and hold the flan together. It is the only one of the three recipes that genuinely reproduces the wobble people are after.

Heat the milk with the vanilla without letting it boil. Beat the eggs with the sugar, pour the hot milk in slowly while whisking, then strain to remove the chalazae. Divide between the ramekins and bake in a bain-marie at 150 °C for 35 to 45 minutes. The flan is done when the centre still wobbles slightly; it finishes setting as it cools. Too hot an oven boils the mixture and produces bubbles and a grainy texture.

Calculated composition per 100 g: 98 kcal, 5.0 g protein, 12.5 g carbohydrate of which 12.4 g sugars, 3.1 g fat of which 1.3 g saturates, 99 mg calcium. Per 165 g ramekin: 161 kcal, 20.5 g total sugars of which 15 g are free sugars from the added sugar, and 163 mg calcium.

Those 15 g of free sugars are half the 30 g daily cap the UK guidelines set for an adult, and 79 % of the 19 g cap for a child aged 4 to 6. This is the point most flan recipes quietly skip. Dropping to 60 g of sugar for the same quantity of milk brings the ramekin down to 10 g of free sugars and remains perfectly pleasant, with the vanilla carrying part of the perceived sweetness.

Sugar: what “homemade” actually changes

The NHS defines free sugars as sugars added to food and drink, plus those naturally present in honey, syrups, and unsweetened fruit and vegetable juices, smoothies and purées. Sugar naturally present in milk, whole fruit and vegetables does not count. That distinction has a direct consequence in baking: if you sweeten a cream with date purée or blended apple compote, you are adding free sugars in the strict sense of the definition.

The caps are numerical: 30 g a day for adults, 24 g from age 7 to 10, 19 g from 4 to 6, 14 g from 2 to 3, 10 g at age one, and no added sugar at all before one. The NHS also notes that the reference intake for total sugars is 90 g a day, of which 30 g free sugars.

What batch cooking genuinely changes here is modest but real: you know the exact weight of sugar you poured in, and you can cut it in 10 g steps from one batch to the next until you find your own threshold. A shop-bought pot gives you no such dial. The idea that a homemade preparation is mechanically less sugary does not survive the arithmetic, though: the panna cotta above, at 60 g of sugar for 800 g of liquid, comes to 8.1 g of sugars per 100 g against 14.7 g for the industrial version, while the flan at 90 g of sugar comes to 12.4 g against 16.0 g. The difference is in the dose, not the kitchen.

The same reasoning applies to syrups and alternatives. Agave syrup shows 70.2 g of sugars per 100 g and maple syrup 62.6 g in the Ciqual table, against 99.8 g for white sugar. Weight for weight they deliver less sugar, but their lower sweetening power pushes you to use more, and the NHS classes them squarely as free sugars. We go into this in more depth in our piece on natural sugar alternatives and glycaemic index.

Storage: the real durations, and why

This is where batch-cooking recipes are vaguest. “Three to four days in the fridge” is a widespread claim and rarely justified.

Food Standards Agency guidance sets the frame. The fridge should sit between 0 and 5 °C, the freezer around -18 °C. Hot preparations do not go straight into the cold: let them cool at room temperature and refrigerate within one to two hours, dividing into smaller portions to speed the cooling. Leftovers should be eaten within 48 hours, or frozen if that is not going to happen. The danger zone runs from 8 to 63 °C.

For these desserts that translates as follows. A milk or plant-drink dessert, poured hot into individual pots, cooled in air and refrigerated within two hours, should be eaten within two to three days. Dividing into small pots rather than one large dish is not merely convenient: it splits the mass and shortens the time spent in the danger zone. The egg flan, which contains coagulated egg and no heat-stable gelling agent, keeps the shortest time; eat it within 48 hours.

The NHS points out that listeriosis is most often contracted from chilled ready-to-eat foods, and recommends keeping the fridge at 5 °C or below and eating such foods within four hours of taking them out. For pots carried to work, that means a cool bag, not a rucksack left in the sun.

A word on freezing. It works badly here. Cornflour creams split on thawing because the starch network releases its water, and an agar gel turns grainy. This is a texture problem rather than a safety one, and there is no satisfying domestic workaround. If you want something for long keeping, look instead at preparations designed for it, such as jellies and jams without gelatin.

Safety: the three points that actually matter

Eggs are the main subject. In a flan baked in a bain-marie they are heated long enough to pose no problem at all. The risk lies in recipes where the egg stays raw or barely warmed, such as some mousses. Thermal resistance data published in 1995 on liquid egg yolk give useful orders of magnitude: at 61.1 °C a 7-log reduction of Salmonella takes 1.4 to 2.4 minutes, while the same reduction for Listeria monocytogenes takes 4.9 to 16.1 minutes. The schedule that suffices for Salmonella does not suffice for Listeria, and adding sugar or salt raises thermal resistance further because it lowers water activity.

NHS pregnancy guidance is easier to apply than those curves. Hen eggs carrying the British Lion stamp, or produced under the Laid in Britain scheme, may be eaten raw or partially cooked; other hen eggs should only be eaten well cooked, and all duck, goose and quail eggs must be well cooked. Since these certification schemes are British and have no identical counterpart elsewhere, the prudent rule outside the UK is to cook the eggs in any dessert destined for someone who is pregnant, very young or immunocompromised.

The second point is the cold chain, covered above. The third is labelling. Date each pot when you close it, not when you take it out of the fridge. A batch session produces a dozen identical pots, and after two days the order of consumption becomes impossible to reconstruct.

The mistakes that keep recurring

Agar added without boiling. Easily the most frequent. The dessert stays liquid and there is no rescue short of returning everything to the boil. If it happens, tip the mixture back into the pan, boil hard for a minute and remould.

Lemon juice or acidic fruit purée poured into the boiling liquid. Acid hydrolysis degrades agar below pH 5.5 and the gel either fails to set or sets then releases. Add the acid after boiling, or build a separate layer.

Cornflour tipped dry into hot liquid. Guaranteed lumps, and they will not dissolve. Always slake it cold.

Judging texture while hot. An agar gel stays liquid until it has cooled towards 40 °C, and a cornflour cream thickens a great deal further as it chills. Correcting the dose mid-cooking on the basis of hot texture leads reliably to over-firm desserts. To test a dose without sacrificing the batch, drop a spoonful onto a cold plate: it sets in two minutes and shows you the final result.

Piling in more cornflour to rescue a soft texture. The 2025 pudding study shows that raising starch from 1 to 5 % lowers gel strength and increases syneresis in storage. If the cream is too runny it is almost always a boiling-time problem, not a quantity problem.

The single large dish. It cools slowly, spends longer between 8 and 63 °C, and gets recontaminated every time a spoon goes into it. Individual pots fix all three.

Adapting to different situations

Without dairy or eggs, stick to the first two recipes and be aware you lose the milk calcium. The egg flan supplies 163 mg of calcium per ramekin; the chocolate cream on fortified soya drink reaches 129 mg, because the drink is fortified to 98 mg per 100 g against 12 mg for an unfortified soya drink. The difference between a fortified and an unfortified plant drink is therefore roughly a factor of eight on this measure. We cover the question in our article on calcium without dairy.

For children the parameter to watch is free sugar, not fat. A ramekin of flan at 15 g of free sugars already covers 79 % of the daily cap for a child aged 4 to 6. Cutting to 60 g of sugar per litre is the single most useful change, ahead of anything else.

For someone who is pregnant, cook the eggs and hold to the 48 hours and 5 °C without any margin.

If you are chasing the texture closest to gelatin, agar alone will not get you there: it gives a brittle gel. Agar plus locust bean gum, or agar plus xanthan, comes considerably closer on the springiness and syneresis figures quoted above. A different route entirely is not to gel at all and to borrow the texture of an ingredient: blended silken tofu makes creams and flans with no added agent whatsoever.

If you prefer whipped preparations to gelled ones, aquafaba follows foam logic rather than gel logic, with even tighter storage constraints.

Running the session

A shared base saves real time. Heat the litre of plant drink with the sugar, take off half for the cornflour cream and use the other half for a second cream flavoured differently. Chocolate, vanilla, coffee and citrus zest go into the individual pan without changing the base.

Do not mix the logics in one pan, though. An agar preparation must boil, an egg preparation must not boil, and a cornflour preparation must boil briefly then be stirred without pause. Those are three incompatible cooking behaviours, and trying to run them together is the surest way to ruin all three.

The efficient order is this: the baked flan first, because its cooking is long and needs no attention; the panna cotta next, because it must be poured immediately after boiling; the cornflour cream last, because it needs continuous presence. Cooling happens in air with the pots open, and lids go on as they enter the fridge, to avoid condensation dripping back onto the surface.

The choice of liquid base is worth a moment. A shop-bought plant drink often contains a stabiliser, and reading the additives in industrial plant milks heads off nasty textural surprises: carrageenan already present in the drink changes how the gel you add behaves. A homemade plant drink avoids that but is less heat-stable and can split at the boil, which is awkward precisely for agar recipes. On chocolate, our vegan chocolate guide covers how to pick bars by cocoa content, which markedly changes the sugar load of the finished dessert.

The same interchangeable-base method works on the savoury side, as in our batch cooking of soups, and the principle of weighed doses applies to agar-agar in general.

The limits of this article

The compositions given are sums of ingredients calculated from Ciqual, not analyses of the finished product. Cooking evaporates water, which slightly concentrates the per-100 g values by an amount we have not measured. The per-portion figures are therefore more reliable than the per-100 g ones.

The storage times quoted come from general guidance for chilled leftovers, not from a challenge test run on these specific recipes. No microbiological trial was carried out on these preparations. The durations given are conservative by construction, and nothing here licenses extending them.

The data on agar and carrageenan come from additive evaluation opinions, which deal with industrial uses and population-average exposures. They describe the chemistry and general safety correctly, but they are not culinary trials.

Finally, the free-sugar and saturated-fat thresholds quoted are British public health reference points. Swiss recommendations also exist and are worded differently, but the orders of magnitude are close and the British figures have the advantage of being published in a directly numerical, checkable form.

Frequently asked questions

How much agar for a litre of liquid? Between 2 and 4 g depending on the firmness you want, so 0.2 to 0.4 %. Weigh it rather than counting spoons: the difference between a level and a heaped teaspoon doubles the dose.

Why did my panna cotta not set? Nine times out of ten the boil was insufficient. Agar only dissolves from around 85 °C, and the simmer has to be held for one to two minutes after it goes in.

Can cornflour replace agar in a panna cotta? No. Cornflour thickens but does not form a gel you can turn out. You will get a thick cream, not a panna cotta.

How long do these desserts keep? Two to three days in the fridge between 0 and 5 °C for the egg-free versions, 48 hours for the egg flan, provided they went into the cold within two hours of being made.

Can they be frozen? Technically yes, but the texture degrades noticeably: cornflour creams weep and agar gels turn grainy.

Is a homemade cream less sugary than a shop one? Only if you put in less sugar. At conventional doses the values are close: the flan weighed here comes to 12.4 g of sugars per 100 g against 16.0 g for a chilled egg flan.

Is agave syrup a better choice than white sugar? Not in public health terms. The NHS classes syrups as free sugars, exactly like sugar.

What do I do with a cream that has wept water after three days? That is syneresis, caused by starch retrogradation, and it is normal and harmless. Whisk vigorously to reincorporate the liquid, or eat the next batch sooner.

Sources