1. Mixing
![]() |
Vitamix |
2. Cooking
Cooking methods and tools
![]() |
Anova immersion circulator in a water bath |
To help get all the mix up to temperature, I like to pull the bag out of the water bath to agitate it a couple of times—5 minutes and 15 minutes after starting cooking.
If these ideas are unfamiliar, please see our sous-vide series.
![]() |
Lab hotplate with magnetic stirrer and thermocouple probe |
![]() |
Thermomix |
![]() |
Kitchenaid heated bowl |
![]() |
A spatula with a built-in thermometer might help on the stovetop. |
![]() |
If you need to scale to commercial quantities, a pasteurizer, like this one by Jaybee, with attached chilling unit, will likely be in your future. It has a built-in agitator, and makes a hardcopy time/temperature graph. Save your pennies. |
If you choose a saucepan on the stove, just be aware that precise temperature control won’t be possible, and that you’ll be losing water to evaporation. Some people make use of this evaporation in order concentrate milk solids, but between temperature control, evaporation control, and constant stirring, this gives you a lot to handle. See the note on evaporation, in the Time and Temperature section.
Time, Temperature, and Proteins
We used to be taught that custard sets at a precise temperature, and that the only variable was the concentration of yolks. We now know that viscosity rises on a continuum, not just with yolk concentration but with temperature2. Custard forms, at varying thicknesses, between 70°C / 158°F and 88°C / 190°F. Additionally, at higher yolk concentrations and higher temperatures (above 75°C), yolks begin to produce hydrogen sulfide gas, and can develop unpleasant flavors. These flavors may intensify with prolonged cooking—which isn’t done during traditional custard making, but we’re going to explore its benefits.
The Chemistry
Food scientists have known for a long time that milk proteins turn into more powerful emulsifiers and stabilizers, just from being cooked to the right degree3,4. During cooking, whey protein molecules (which make up around 25% of the total milk proteins) begin to unfold (denature), exposing reactive surfaces that had previously been concealed. These are able to form new bonds with casein molecules (which comprise most of the remaining protein). In doing so, they work in similar ways to the emulsifying ingredients we traditionally include in ice cream recipes (like egg yolk). The denatured whey protein can augment, and in some cases replace, traditional emulsifiers. See the post on Emulsifiers.
The partially denatured whey proteins can also form aggregates with other whey proteins, and form a weak hydrocolloid gel—much like the stabilizing ingredients. With careful management, the cooked proteins can serve most of the purposes of an egg custard.
Fortunately, some ice cream pioneers have done the empirical research for us. Engineers at Haagen Dazs aren’t sharing, but Jeni Britton-Bauer, of Jeni’s Splendid Ice Creams, devised a process that includes pasteurizing her egg-free mix at 75°C for up to an hour. Ice cream mix pasteurizes to federal standards in about two minutes at this temperature; the rest of that cooking time works the proteins. Between using a higher than normal level of nonfat mix solids, a long, controlled cook, and a small addition of relatively weak starch stabilizer, she creates egg-free ice creams that have a rich mouthfeel normally associated with custard bases.
Flavor
Britton-Bauer also favors this time / temperature combination because it gives a custardy “cooked milk” flavor to the mix, which she favors. My own experiments don’t bear this out. I made identical, unflavored samples of ice cream mix (15% fat, 10% nonfat milk solids, 2 yolks/liter), cooked sous-vide at temperatures from 72°C to 80°C, for times ranging from 15 to 60 minutes. In a series of double-blind triangle tastings, subjects found the differences almost too close to call. There was a slight preference for 75°C held for 30 minutes, but there was no sense of cooked milk flavor, and it was not more “custardy” than others. It’s likely that with flavoring ingredients added, there would be no differences. I take this as good news: at least if your mix is low on egg yolks you shouldn’t have to worry about flavor when choosing a cooking time and temperature.
You may want to run your own tests. Variables like the number of yolks and the level of nonfat milk solids might influence the results.
If you’re cooking the mix for more than a few minutes in an open container (every method mentioned here besides the immersion circulator, Thermomix, or the hot plate with a covered container), enough water will evaporate to effect the results. You’ll see a decrease in the total mass of the recipe, which will correspond to an increase in the percentage of solids. This will improve most cookbook recipes. But it can hurt a well designed professional recipe that already has optimal solids.
If you’re doing long cooks in open containers, weigh the mix again after cooking, and recalculate your formula based on the final mass. You’ll find you can add less dry milk.
Some people use evaporation deliberately to increase nonfat milk solids. This can work, of course, but I’ll suggest that you’re sacrificing precision by juggling too many variables at once. Evaporation rate will vary with the size and shape of the cooking container, the volume of ice cream mix, and even the ambient temperature and humidity. How are you going to control both the evaporation and the cooking of the milk proteins?
Some people chose evaporation over dry milk powder on the assumption that dry milk powder tastes bad. Some of it does; the good stuff doesn’t. Higher quality versions are pure, spray-dried skim milk, with the water evaporated at lower temperatures and under more precise controls than you’ll create in the kitchen. You’re doing yourself and your ice cream no favors by taking on this role.
Time, Temperature, and Protein Conclusions
Hydration
Pasteurization
3. Homogenizing
![]() |
A rotor-stator homogenizer generates higher shear forces than any blender |
Homogenizing means blending with such high shear forces that the big fat globules are broken up into many much smaller ones. This improves smoothness, whipability, and stability of the foam structure.
Even if you start with homogenized milk, it helps to homogenize again after cooking. At cooking temperatures, the fat globules melt and start glomming onto one another. Meaning, the milk will start to de-homogenize.
4. Chilling
5. Aging
6. Spinning (Freezing)
![]() |
Carpigiani Maestro and Labotronic machines, along with similar models by Bravo, are about as good as they come. |
![]() |
Fill one side with your mix, the other with salt and ice, and give to some ADHD kid to kick around. |
![]() |
Ice cream attachment for mixer. Works well if your freezer’s set to -5°F or lower. |
![]() |
Lello Musso gelato machine. Among the best compressor-driven home models. |
Filling the Machine
If your mix has been aging in a Ziploc bag, the easiest way to get it into the machine is to snip off one of the bottom corners and use the bag like a pastry bag: squeeze the mix out. Some mixes can be quite viscous, or even gelatinous, especially if there are lots of added fats from chocolate or nut butters, or if your stabilizer blend forms a gel (which should be avoided).
If your mix has formed a gel that’s too viscous for the machine to churn efficiently, squeeze the mix into a bowl or plastic container, and blast it with a stick blender. This will turn the stiff gel into a fluid gel that’s much more pourable.
Drawing Temperature
It’s good practice to aim for a consistent drawing temperature; a temperature when the consistency is correct and it’s time to remove the ice cream and head to the next step. It’s tempting to just look at texture and ignore the temperature, but this won’t let you fine tune your mix. Here’s how to do it:
Use an instant read thermometer (a Thermopen or equivalent) to check the temperature of the ice cream when the consistency looks right—dry on the surface, and relatively smooth, with a viscous, whipped, soft-serve texture). At this point the ice cream has achieved the right structure. In the interests of ideal consistency at serving temperature, and in limiting ice crystal growth, we want the mix to reach this consistency at -5°C / 23°F to -6°C / 21°F.
If the ice cream measures warmer than this when the consistency is right, the recipe should be adjusted to increase freezing point suppression. And of course if it’s colder, the recipe should be adjusted the other way.
If your machine has the cooling power to freeze the ice cream colder that -5°C, you may find it advantageous to let it do so—provided that the rate of cooling is still faster than that of the freezer you use for the hardening step. If it take your machine an additional 20 minutes to shave off a few more degrees, then waiting will likely hurt the texture. It’s probably not practical to go significantly lower than -6°C / 21°F. The process becomes self-limiting; the ice cream is stiff enough that the machine will be adding heat in the form of mechanical energy as fast as the compressor can remove it. Thrashing the ice cream around at a steady temperature is counterproductive.
Alternative Freezing Techniques
![]() |
Meet Paco. |
The Paco Jet starts with mix that’s been frozen solid; it uses high-speed blades to shave and fluff the ice into a foam with nano-sized ice crystals. The ice cream is usually made to order at small cafes and restaurants, so recipes can be looser and freer than ones designed for conventional machines. A smooth texture is guaranteed, provided someone eats the ice cream right away. If you need to make the ice cream ahead and store it in a freezer, then all the usual considerations apply again.
![]() |
Ahhh, the cryo life. |
Liquid Nitrogen came onto the scene in the 1980s, possibly discovered by the founder of Dippin’ Dots as a viable ice cream freezing method. LN2 boils at -196°C / -321°F, so as you might imagine, it freezes ice cream fast. This means almost preternaturally smooth ice cream, which can be made in just a couple of minutes in an ordinary stand mixer.
That is, if you have a laboratory dewar full of liquid nitrogen handy. You can buy a dewar for several hundred dollars, and you can probably find a service that will deliver to your kitchen. The expense and inconvenience may still be discouraging. LN2 isn’t terribly expensive, but you need a lot—several times the volume of your mix. And delivery services may have minimums. So LN2 ice cream has mostly been popular with people who have other reasons to keep liquid gas around, like chefs who work with contemporary techniques, or biologists who like to party.
To make LN2 ice cream, wear the appropriate safety gear—especially shoes or boots that are closed and covered by your pants cuffs. Make sure there’s good ventilation (otherwise, if you spill a bunch, nitrogen will displace the oxygen from the room and you’ll never know what hit you). Pour your mix into a stand mixer work bowl, turn the machine on low, using the flat beater, and slowly pour the LN2. Enjoy the disco fog. Stop pouring and stop the machine when you get the consistency of soft serve. Don’t try to over-harden it … your mixer won’t be happy.
Dry Ice offers some similar advantages to LN2. With a melting / sublimation point of -79°C / 109°F, it’s positively balmy compared to nitrogen, but it’s a lot colder than your ice cream machine. Dry Ice isn’t as easy to find as it once was—those gel-filled cold packs actually do a better job of keeping beer chilled in the cooler, and won’t burn anyone. If you can find some locally, make sure it’s food grade. A lot of the stuff is technical grade and can be contaminated with god-knows-what.
A unique quirk of dry ice: it will carbonate your ice cream. The C02 bubbles are quite unstable, so the carbonation won’t last long. If you don’t want it, hold the ice cream in your freezer for a couple of days. If you do want it, serve the ice cream as soon after hardening as possible. I’ll post my recipe for sparkling champaign sorbet in a later post.
To use dry ice, grind it into snow or small chunks in a blender, and then churn into the mix with a stand mixer, as with LN2.
7. Hardening (Static Freezing)
![]() |
Blast freezer! |
Most ice cream machines will only get the ice cream down to soft-serve temperatures. You then need to harden the ice cream, and do so as fast as possible—again, to keep the ice crystals from getting too big. Crystals grow rapidly at the relatively warm drawing temperature—the clock is ticking as soon as stop the ice cream machine.
Commercial ice cream kitchens and better equipped pastry kitchens use blast freezers, typicallly set to -40°. If you only have a standard freezer, the tricks to a quick freeze are to set the freezer as low as possible (most newer freezers can get down to —20°C / -5°F, and to freeze the ice cream in small batches. Pints are ideal. 16oz polypropylene takeout containers are especially convenient. They get brittle in the freezer, so they crack easily, but you can always stock up by ordering some more takeout.
If the ice cream won’t be eaten immediately, you can add some insurance against iciness and off-flavors (from the freezer) by pressing a layer of plastic wrap over the surface, before snapping the lid on. Be sure to use odor-free plastic wrap (made from polyethylene) rather than the commercial types (made from polyvinyl chloride). All the consumer brands are ok. The commercial brands form a better oxygen barrier, but will make the ice cream taste like a shower curtain.
![]() |
Lots of these. |
If you’re using a blast freezer, you’ll wait until the ice cream is hardened throughout, and then transfer to a standard freezer. It will be as hard as concrete, and will take many hours to resemble dessert. If you’re using a standard freezer for hardening, then after 8 hours or so you’re done. Moving the ice cream to the fridge 15 minutes or so before serving will make it more scoopable. Ideal serving temperature is 6 to 10° F / -14 to -12°C.
Closing Thoughts
In the next post we’ll take a closer look at sugars.
Also see Goff, H. Douglas. “Partial Coalescence and Structure Formation in Dairy Emulsions,“ in Food Proteins and Lipids, Volume 415 of the series Advances in Experimental Medicine and Biology pp 137-148
How would you add yogurt to a recipe? I’m afraid cooking it with the base would cause the whey to separate and curdle. Should it be added once the rest of the base is cooked and then chilled?
Hi Ryan,
I haven’t tried these methods with yogurt, but suspect there won’t be any problems. The cooking temperatures are lower than what typically causes yogurt to curdle, and the presence of stabilizers makes the yogurt matrix stronger. In Indian cooking, starches are often used in part as a stabilizer when yogurt is going to be cooked in a sauce. In ice cream, the stabilizing ingredients (whether gums, starches, or proteins) should work similarly. Full-fat yogurt is much more stable than low- or no-fat, so it’s the better choice.
Would there be any considerations in recipe construction when using a typical soft-serve machine, or should I just treat it like any other ice cream machine?
Soft serve formula are typically very low fat, quite sweet, are egg-free, and have a bit less freezing point depression than regular ice cream (because serving temperature is warmer).
We’ve formulated one soft serve formula for a client, and tried to make something more satisfying than the typical industrial blend. We went for similar sweetness levels to the regular ice cream they make, and 10% fat, and blended the sugars to give a few percent less freezing point depression. Total solids was around 36%.
Regarding protein denaturing, I wonder if this really makes that much of a difference in homemade ice cream. From what I’ve found, to denature 100% of the whey protein requires boiling milk (212 F°/ 100 C° ) for 10 minutes.
Boiling milk for ice cream doesn’t seem like a good idea (it will cause a cooked dairy taste which in some recipes is not a bad idea).
The reason I’m asking is that I’d like to make custard-based ice cream without cooking it. I’d pasteurize the eggs and then I’d be done with the stove.
Thanks,
Thanks for writing. You’re right to wonder how much difference this makes in homemade ice cream, but not for the reasons you’re thinking.
The goal isn’t to fully denature 100% of the whey protein … just as the goal when cooking steak isn’t to turn the meat into a uniformly gray, dried-out dog chew toy.
We’re trying to denature the proteins to the right degree—the level of cooking that unwinds them to the point where just enough oleophilic molecules are exposed so the proteins can do our bidding.
If the whey is overcooked, you risk the whey molecules unwinding to the point where they form an affinity for each other. They can then form aggregates that alter the texture in different ways. I have zero experience with this, but studies indicate that when cooked past the point where they function as emulsifiers, they can reach a point where they will function as a stabilizer, and beyond this, to a point where they will cause gritty texture problems.
You also can expect cooked milk flavors, but I don’t know how far you’d have to go for this to be noticeable. Dairy flavors in general have a hard time competing with the sugar and the flavor ingredients.
There are a few reasons that whey denaturization might have little effect in homemade ice cream:
• There usually isn’t a lot of whey. Our formulas have a lot of whey; that’s where the added skim milk solids come in.
• The whey is often already denatured too far. This happens when you use ultrapasteurized dairy, or skim milk powder that’s been spray-dried at high temperature.
• Even with low-temperature pasteurized dairy, you don’t really know what’s going on with those proteins. Some larger scale ice cream makers who care about this stuff start with raw milk, and precisely control the pasteurization process. Jen’s Splendid is an example. I assume Haagen Dazs is also.
• Lots of homemade ice creams are so high in fat and egg custard that you’d never know the difference; they just don’t need any subtle help from the whey proteins. There’s a website presumably about ice cream science that talks for pages about the importance of denaturization …. and then gives recipes that are so high in fat, solids, and egg custard that you could never in a million years tell the difference.
In our formulas, we either de-emphasize egg yolk or eliminate it, depending on the type of flavor. We’re precise about nonfat milk solids levels. We urge you to use low-temperature pasteurized milk and cream, and low-temperature spray-dried milk powder. Our fat levels, solids levels, and stabilizer levels are carefully considered, but not extreme.
Still, I’d urge you to do a side-by side test, especially if you have a sous-vide setup or other arrangement that gives you temperature control. Let us know how it goes.
Thanks for giving me a clue!
I think the website you’re referring to is ice cream science. And yes, his (Ruben) ice creams are often well above 20% butterfat!
What got me started down this path is Dana Cree’s book, “Hello My Name is Ice Cream.”
I have a sous vide, as well as DIY PID controller/temperature probe setup that works well to control the temp on my Cadco hot plate.
Per the ice cream science web site, “Sava et al. (2005) note that thermal denaturation of whey protein involves 2 steps: an unfolding step at 70 to 75°C (158 to 167°F), and an aggregation step at 78 to 82.5°C (172.4 to 180.5°F), that mostly follows unfolding. Phillips et al. (1990) note that foaming and emulsifying characteristics may be impaired if protein undergoes aggregation.”
So what I’ll do is sous vide my milk (I’ll use one that’s not ultra pasteurized) and keep it at 167°F for 25 minutes and then make a Philadelphia style ice cream: 2 cups cream to 1 cup milk and 3/4 cup sugar. Since there’s no SMP or stabilizer (guar gum would be my choice) I’m interested in know if just the whey protein in the milk is enough to make a difference. I have my doubts.
I let you know how it goes.
From icecreamscience.com:
TIP#5 – 72°C (162°F) FOR 25 MINUTESI’ve found that when I hold my mix at temperatures above 72°C (162°F) for a prolonged period of time, the unpleasant ‘eggy’ hydrogen sulphide taste begins to form and becomes noticeable on eating. I would therefore recommend heating your ice cream mix to 72°C (162°F) and holding it at this temperature for 25 minutes as this significantly improves ice cream body and texture. I’ve run several tests where I have kept the temperature constant at 72°C (162°F), as well as the composition, but have varied the heating times. I’ve found that a mix heated for 25 minutes produces smoother and creamier ice cream than compared to mix heated for 5, 10, and 15 minutes at the same temperature.
It’s important to look at the specifics of the formula in question before generalizing from results like this. For our most basic base formula (15% milk fat, 12% msnf, 2 yolks/liter) we did several rounds of blind triangle tests with different combinations of temperature and time (unflavored base).
We could not discern a flavor difference, even between the extremes. There was never a hint of egg yolk or sulfides.
The unscientific component in our methodology is the difficulty in predicting time/temperature curves of liquid product heated in a water bath. We’ve been using 77°C, primarily as a way to guarantee that at its peak temperature, the base is hot enough to hydrate the locust bean gum.
The instructions you quote are from a source that has not discussed the impact of its approach to formulation (very high fat, very high solids, very high yolk content).
You’re amazing! I love the science explanation behind The Ice Cream. I have a question concerning sweetened condensed milk as an ice cream base. Recently, I’ve made vanilla ice cream using 1 can of sweetened condensed milk ( 14 oz) and 16 oz of heavy cream 5% fat). No eggs and no other stabilizers were added. I also didn’t heat up the any of the ingredients. The results was really good. Very creamy, smooth, and silky; it left a nice feel on my palate. So, is it the slow process of sweetened condensed milk being cooked for a long time, possibly produced it’s stabilizer? A lot of people are giving out this recipe but no one seems to be explaining how or why it works so well.
Hi Paulette,
That could be the explanation. It’s also likely that you’re getting the benefit from a high milk solids content.
Please, don’t ridicule people with ADHD. Your article about ice cream technique is good without those kinds of unnecessary remarks.
I ridicule no one. If I didn’t have ADHD myself, I’d be doing what I’m supposed to be doing instead of blogging about ice cream.
Hi,
I was thinking that a high speed blender could be used both to homogenize and heat the base to around 75 C. Would you advice for or against?
You could probably do this. But a blender is a very inefficient heater—it would make for a long and loud and unpleasant process. And your blender motor would not thank you.
Apparently several companies have created “cooking” blenders. Ninja and KitchenAid are two. Ninja has demo’s on YouTube. I have never actually seen one, much less tested, so I’ll have to wait to until some rich, brave soul breaches the ramparts.
Hi,
I’m looking to purchase a batch freezer from Carpigiani but they all seem very similar in terms of performance. Is there a specific model you would recommend? I will be making ice cream, not gelato.
Thank you!
Probably best answered by a Carpigiani rep. I doubt you’ll find many people who have used multiple current models of their machines.
In regards to denaturing the milk proteins to increase emulsification/stabilization ability…
It reminds me of the same process in Cheesemaking; Specifically when making Mozzarella and heating the curds while stretching them; which iirc takes place around 75-80 c also.
Thank you for this useful article.
This is a great article! The only one of its kind on the internet right now.
About homogenizers: I’m having a hard time determining the qualitative differences between a $180 stick homogenizer (e.g. the Fama Gastro-Inox 505.175) and a $2000 one (e.g. the pro250 laboratory-grade homogenizer in your example). I also have no clue where the relative performance of a high-powered (Vitamix) blender would fall in this comparison. There just seems to be no information available online.
Any thoughts?
First, thank you for bringing the Fama Gastro to my attention. I was not aware of any homogenizers being made specifically for kitchen use. This does not seem to be available outside of Europe. At first glance I didn’t even see an English language version of an online store that sells it.
I would not assume that a product like this is inferior to the lab models in any important way, despite the 10X price difference. Lab stuff is just expensive. Remember when the cheapest immersion circulators were the $1,100 PolyScience lab models. These are probably more reliable and heavy-duty than the typical $150 circulators you can buy today, but the differences aren’t important to most of us most of the time. There are actually advantages to the cheaper ones (size, weight, etc.) But to know for sure you’ll have to wait for competent reviews.
The amount of shear you get will depend on the motor speed / power, and the design of the rotor and stator. I see no reason that this couldn’t be done adequately for under $200. A friend of mine has a lab-grade homogenizer, and says that its motor is actually made by Bamix!
If you look at lab homogenizers, be sure to check the capacity. Many labs have no need to homogenize more than a few ml of solution in the bottom of a test tube; lots of the more affordable lab homogenizers are designed for these quantities and would be useless in the kitchen.
In general, the amount of shear you can expect is, from lowest to highest: regular blender < high-shear blender (vitamix etc.) < rotor/stator homogenizer < ultrasonic homogenizer < high-pressure homogenizer (like what they’d use at a dairy or ice cream manufacturing plant). I don’t know how to quantify the differences. I do know that there’s a significant difference in appearance / texture / stability of emulsions when you move from a regular blender to a high-shear one, and again when you move to a rotor/stator homogenizer.
I couldn’t speak to the difference between different power levels of rotor/stator homogenizers. Or to the more powerful classes of homogenizer, which aren’t really options for our purposes.
Edit: that should be a ~$600 Fama stick homogenizer. I realized after writing the comment that $180 is just the stick, and the motor costs another ~$400.
The cheapest pre-combined solution I could find was actually the Misceo 250P, at $500.
I’m buying an Anova due to this page. You should ask for a referral code for your Underbelly peeps. I’m sure you would have made a trillion dollars in referrals by now
Magimix Cook type devices for mixing icecream ingredients/denaturing/pasteurizing are an absolute disaster. I bought an 1100,- model three years ago and it was completely incapable of doing what’s needed for creating ice cream mixes. The blades are higher in the bowl facing upwards and that means all the sugar will just stay at the bottom of the bowl under the blades, caramelize and not mix at all. Also, while the temperature can be set to within a degree of accuracy on the display, the machine does not allow you to read out the temperature. So all cold ingredients are in the bowl but if you set the timer to 10 minutes at 75 degrees, the timer will just start right away, not when it reaches temperature. Further more, there is no ‘slow heat to set temp’ feature. It will just fully heat until the set temperature is reached, thats not delicate enough for heating milk. The device is perfect for what it’s designed for but it is not designed to create ice cream mixes.
Good to know, thanks. There are surprisingly few good solutions for small-scale cooking.
Fantastic article, thank you