The short answer: First pour from a few inches up into the centre, so the milk dives under the crema, then lower the jug close to the surface and pour more slowly, so the foam stays white on top. That is how Sage UK describes it. The physics behind it is an inverted fountain: more pour height means more mixing and brown foam, gentle pouring gives white (Mathijssen et al., Reviews of Modern Physics, 2023).
Frothing milk and pouring milk are two separate skills. The first decides whether there is pourable foam in the jug, the second whether it becomes a pattern. Frothing itself is covered in the guide How to froth milk; this one is about the way from the jug into the cup.
Step by step: your first heart
Sage UK divides pouring into three phases, foundation, shape and finish, and names the heart as the starting point for new baristas. Breville, Sage’s sister brand, puts one more step before it in its tutorials for the Bambino Plus: the heart begins as the dot and only changes the technique at the end. The steps below follow Sage, with the reason given for each.
- Prepare the foam. Tap the jug briefly after frothing and swirl the milk until the surface is evenly glossy. Sage UK names tapping and swirling as the remedy for large bubbles; Lelit has you turn the jug after tapping.
- Pour the foundation. Pour a steady stream from a few inches up into the middle of the espresso. It breaks through the crema and mixes the milk in beneath the surface. Everything on top stays brown, which is intended here.
- Lower and slow down. As the cup fills, bring the spout closer to the surface and pour more slowly. Now the foam no longer dives under, and a white dot appears in the middle.
- Build the white. Keep pouring at the same spot and the dot grows into a round patch. The jug stays still; for the heart you need no sideways movement.
- Pull through. At the end, lift the jug and draw the stream through the middle of the patch in a single, calm motion. The thinner stream, now falling from higher again, cuts into the circle and pulls it out into the tip of the heart.
Sage does not tie the moment of lowering to a fill level. The blog says the jug moves closer as the cup fills, for the heart as soon as a white dot appears. The common mark “when the cup is half full” turned up in no manufacturer or specialist source we found. We still consider it usable as a guide for the first attempts, because enough room is then left for the pattern.
Why pour height decides everything
Because it controls whether the milk ends up under the crema or on top of it. Mathijssen, Lisicki, Prakash and Mossige, in their review of fluid mechanics in the kitchen (Reviews of Modern Physics, 2023), classify the pouring of milk foam into coffee as an inverted fountain. This means a jet pushed into a heavier liquid while buoyancy works against it. How deep it penetrates and how much it entrains depends, according to the paper, on the jet diameter and the pour height, summarised in the Froude number. According to the authors, large fountains lead to stronger mixing and brown foam, while gentle pouring leads to white.
This gives the rule “first high, then low” a physical basis. From height the stream arrives faster, penetrates deeper, and the milk disappears under the crema. Close to the surface it lacks that speed, the lighter foam stays on top and becomes visible. The authors also write that after a lot of practice both colours can be combined in quick succession into patterns such as the heart and the phoenix.
For strong fountains they give, after Turner, an approximation: the penetration height grows roughly linearly with the jet radius times the Froude number (z_f ≈ 2.46 · L0 · Fr), which recent experiments confirm well. Other scalings apply to weak fountains. Applied to the jug, this means a thick, fast stream dives deeper than a thin, slow one. Anyone who also throttles the flow while lowering, as Sage describes, reduces both. This is our derivation from the model; we know of no measurement on milk foam. For miscible cocktail layers the authors advise slow pouring for the same reason.
For the pull-through at the end of a heart it works the other way round: lifting the jug gives the stream fall height again, it pierces the white patch and carries brown forward.
What milk foam does latte art need?
Fine microfoam without large bubbles, which flows from the jug instead of clumping. For the steam wand, Sage UK describes five steps: fill the jug halfway with fresh, cold milk, purge the wand briefly, place the tip just below the surface, turn the steam on and tilt the jug slightly so a whirlpool forms. At first the tip may break the surface slightly and draw in air, then it goes deeper and distributes the air. Here are the faults too: according to Sage, large bubbles mean too much air, so next time put the tip deeper. Thin milk means too little air, and the tip belongs higher in the first seconds. A cooked taste comes from steaming too long.
Temperature is one of the figures on which the sources disagree. De’Longhi gives 60 °C as the ideal in the Dedica instructions, Sage 60 to 65 °C in its instructions, and Lelit sets the upper limit at 65 °C. Tchibo and Simpletaste give 60 to 70 °C, Philips for the Milk Twister about 74 °C, and the barista Eric Wolf, in a 2018 conversation with Stiftung Warentest (Germany’s consumer testing foundation), 65 to 70 °C. The Sage latte art blog gives no figure in degrees at all; there you stop when the body of the jug feels uncomfortably hot. What is documented is 65 as the limit of the International Agency for Research on Cancer (IARC, 2016) for very hot drinks, and that concerns drinking temperature. Our assessment: stop at 60 to 65 °C and measure with a thermometer, because the hand check is itself described differently between the Sage sources.
On the choice of milk, two camps face each other. Sage UK recommends whole milk for beginners, because of its creamier texture and stable microfoam, and calls low-fat milk, soy drink and almond drink harder to froth. Foam research, by contrast, sees fat as a hindrance: Huppertz (International Journal of Dairy Technology, 2010), Kamath et al. (2008) and Ho et al. (2024) describe poorer foam formation and more bubble coalescence with more fat. De’Longhi and Philips recommend semi-skimmed milk. The studies measure foam volume and stability, not how well foam can be poured. For learning we would therefore follow Sage and stay with whole milk, and above all not change type until the heart works.
How long may the foam stand? Lelit even has you let the foam rest for at least half a minute after tapping. Ho, Xiong, Bhandari and Bansal (Food and Bioprocess Technology, 2024) found larger bubbles in all samples after 10 minutes, because liquid drains out of the foam and bubbles merge. There is a lot of room between these two statements. We would pour straight after swirling and time the espresso so that it waits, not the milk.
What the crema contributes
The crema is the brown base on which the white becomes visible. According to Illy and Navarini (Food Biophysics, 2011), its gas phase comes from the carbon dioxide of roasting. How much of it reaches the cup is influenced by tamping pressure, the time between grinding and extraction, roast degree, dose, brewing conditions and type of coffee. Good crema should, according to the paper, make up at least 10 per cent of the espresso volume and last at least a few minutes before a first uncovered spot appears.
In practice the ground coffee matters most. Ground coffee stored in the air for a long time also loses foam stability: according to Severini et al., cited in Illy and Navarini, 36 hours were enough for the so-called sugar experiment to give only 0.5 seconds (the range of measurements was 0.5 to 14 seconds). This applies to powder, not to whole beans. Sage UK accordingly warns that with beans that are too old or with pre-ground coffee there may be no crema at all, and advises freshly ground espresso.
Crema does not always disappear completely with old coffee, however. Smrke et al. (2018) explain this by carbon dioxide that stays bound in the coffee matrix under dry conditions and is only released when extracting with water. Pale crema is regarded by Illy and Navarini rather as a sign of under-extraction, more on that under the faults. How to set up a stable shot is covered in the guide Espresso basics.
The numbers at a glance
| Quantity | Value | Source |
|---|---|---|
| Pour height for the foundation | a few inches, no centimetre figure | Sage UK, latte art blog |
| Filling of the jug | at most half | Sage UK; Lelit, Victoria instructions |
| Milk volume when frothing | two to three times, stated for the Dedica’s cappuccino foam, not for pouring microfoam | De’Longhi, Dedica EC685 instructions |
| Milk per cappuccino | 100 ml at about 5 °C | Lelit, Victoria instructions |
| Final milk temperature | 60 °C (De’Longhi) to about 74 °C (Philips) | Manufacturer instructions, see above |
| Resting time after frothing | at least half a minute | Lelit, Victoria instructions |
| Bubble growth in the foam | larger in all samples after 10 minutes | Ho et al. 2024 |
| Crema share | at least 10% of the espresso volume | Illy and Navarini 2011 |
| Ground coffee in the air | after 36 h the sugar experiment gives only 0.5 s | Severini et al., cited in Illy and Navarini 2011 |
Which cup and which jug make it easier?
For the cup there is a manufacturer recommendation: for the tulip, Sage UK writes that a cup with a wide opening and bowl-like shape gives the milk more room to spread. More area means more room for the pattern. For a tall, narrow glass, all that can be derived from the fountain model is that milk falling deeper arrives faster and mixes more.
With the jug, the spout is disputed. Sage UK advises a narrow spout for better control and preferably a smaller jug, so the heat is kept while frothing. How little even manufacturers agree in their descriptions is shown by De’Longhi: the milk jug DLSC060 (350 ml stainless steel, about €16.99) is called “narrow spout to create latte art” on the US site, while on the New Zealand site the same model has a “wide spout”. The NewGF jug (350 ml, 304 stainless steel, scale inside, about €11.04) has, according to the Amazon listing text, a narrow, tapering spout (there is no manufacturer evidence for this). Our assessment: if you are buying new, follow Sage and take the narrow spout; for a first heart without sideways movement, the design matters less than the foam.
The fill level is clearer. Cold milk at most halfway: Sage UK and Lelit agree on this. How large the jug has to be follows from this rule and Lelit’s 100 ml of milk per cappuccino. The table rests solely on these two statements and is our calculation:
| Plan | Milk | Jug, by calculation | Basis |
|---|---|---|---|
| one cappuccino | 100 ml | from 200 ml nominal volume; a 350 ml jug holds 175 ml half full and is enough | Lelit (100 ml), half-full rule (Sage UK, Lelit) |
| two cappuccinos from one frothing batch | 200 ml | from 400 ml nominal volume | 2 × 100 ml, doubled by the half-full rule |
| Comparison values from manufacturers | Sage supplies the Barista Pro with a 480 ml jug, De’Longhi lists the DLSC069 with up to 500 ml alongside the DLSC060 | Sage product page; De’Longhi according to the DLSC060 review |
A jug of 480 or 500 ml is above the calculated lower limit for two cappuccinos. You must not transfer the factor from the Dedica instructions, by which the milk volume rises to two to three times, to this calculation. De’Longhi gives it for the thick cappuccino foam of its own frother, and 175 ml of milk would fill a 350 ml jug or make it overflow. None of the instructions used here gives a figure for how much pourable microfoam grows.
From heart to tulip to rosetta
The sources agree only on the beginning: the heart (with Breville the dot before it) is the entry point. After that the sequences diverge. Sage UK calls the rosetta “a step up in complexity” from the heart and describes the tulip after it as a more elaborate pattern. Breville puts tulip and rosetta together after dot and heart, without a ranking.
| Pattern | Movement according to Sage UK | What can go wrong |
|---|---|---|
| Heart | steadily into the middle until a white dot forms; jug closer, slower; lift and pull through in one calm motion | lowered too late, no white appears |
| Rosetta | start as for the heart; from the white dot, while continuing to pour, gently zigzag sideways; slow down and pull through the middle so a clear stem forms | zigzag without an even flow, the leaves run |
| Tulip | small dot in the middle; stop, pour the next dot directly onto it, adjust the flow; repeat several times, stay central; pull through at the end | poured without stopping, the segments merge |
How often you stop for the tulip, Sage leaves open with “several times”. For the order we have an opinion that follows from the descriptions: the tulip only repeats the movement of the heart with pauses in between, while the rosetta demands a second hand movement, the sideways zigzag, at the same time as the retreat. We would therefore practise the tulip after the heart. Anyone who follows Sage and takes the rosetta first is doing nothing wrong.
Why does my pattern not appear?
No white in the cup is the most obvious picture, and the cause follows directly from the fountain model: the stream had too much height or too much flow until the end, and the foam dived under completely. Lower the jug earlier and pour more slowly as you do.
The white appears and runs? Then either contrast in the base or substance in the foam is missing. Illy and Navarini read a pale crema as a sign of under-extraction: ground too coarse, water too cool or time too short. A very dark crema with a white spot in the middle points, according to the same paper, to ground coffee that is too fine or too much coffee. The authors themselves classify such colour rules as mostly based on experience. You adjust the grind following the guide Coffee grind size. If the crema is fine, the milk is to blame: thin, liquid milk had too little air, according to Sage.
Large bubbles on the surface mean, according to Sage, too much air when frothing. Tapping and swirling partly rescue this jug; for the next one, the tip goes deeper. Lumpy foam that does not flow is, in our assessment, whipped too firmly, so also too much air. No pouring technique helps here; the correction belongs at the steam wand.
A pattern that sits crooked or wanders to the edge probably had an impact point off centre. Sage describes pouring centrally and pulling through the middle for all three patterns. Watch the point where the stream hits the surface and leave the jug there.
A heart without a tip, ending as a round patch, lacks the pull-through according to Sage’s sequence: the jug was not lifted at the end. Only the fall height gives the stream the force to cut into the patch.
Which equipment is suitable for latte art
At competitions the question is settled: the rules of the World Latte Art Championship allow only a hand-operated steam wand for the milk. For the kitchen this is no law, but it is a hint at what professionals expect. The barista Eric Wolf, in an interview with Stiftung Warentest (test magazine 2/2018), calls portafilter machines with a nozzle ideal and frothing a matter of practice; electric frothers, he says, are good for anyone without great ambitions, but nothing for the high art of milk frothing. No measurement stands behind this verdict.
Against this stand manufacturer statements. WMF describes the warm foam programme of the WMF Stelio as warm, creamy and fine-pored, especially for cappuccino or latte art. Severin, in the data sheet, calls the barista pouring spout of the SM 3587 also suitable for latte art, and the Subminimal NanoFoamer Lithium has two exchangeable NanoScreens for different textures. With the LattoGo, OutIn explicitly recommends one setting for latte art. A comparison that measures the pourability of such devices against a steam wand does not exist, according to our research. Our assessment: the steam wand gives you the most control, because you manage air intake and temperature yourself. With the Powerlix Pro 3 in 1 with wire spiral, the manufacturer does not promise latte art at all, and our German review considers it the wrong choice because of the rather loose foam.
Among machines with a wand, the limits differ. On the Sage Barista Express you control air intake and jug angle yourself; the steam function stops after 5 minutes of continuous steam. The Gaggia Classic Evo Pro is a single-boiler machine: steam for at most 60 seconds at a time, after which the boiler must cool down before the next espresso, and no milk jug is included. The Sage Bambino Plus froths with Auto MilQ in three temperature and three foam levels and, according to Sage, also allows manual texturing, but for frothing only in the supplied stainless steel jug, filled between MIN and MAX. For the Sage Creatista Plus, Sage names microfoam for latte art as the purpose of the automatic nozzle. Our German review considers a heart feasible with it; for competition patterns, it says, the freely guided wand is missing.
Fully automatic machines that deliver the milk foam directly into the cup leave, in our assessment, hardly any room for a poured pattern, because the jug is missing.
What no source documents
Some statements often found in latte art instructions we could not trace to any manufacturer or specialist source. We collect them here so you know what you can rely on and where you have to try things out yourself.
A pour height in centimetres is given nowhere; Sage speaks only of a few inches. The often-cited tilted cup at 45 degrees comes from blogs and advice portals. The same goes for practising with water and a drop of washing-up liquid in the jug: a common tip for training the hand guidance without milk, but with no evidence of how well it reproduces the behaviour of microfoam. We know of no study on the learning curve, that is, how many cups are needed until the heart is round. No work we know of has measured latte art itself in terms of fluid mechanics (pour height, jet speed, foam density). The classification by Mathijssen et al. is a review; for the patterns the authors refer to an online guide (Bez, 2021). From the SCA (Specialty Coffee Association), the only freely accessible material on pouring is the 2026 WLAC rules with definitions, cup sizes and scoring; there are no figures on technique in them.
Then there are statements often read that do not hold up. Above 65 °C the protein framework of the foam supposedly becomes unstable and it collapses faster: none of the instructions cited above names such a tipping point, and the 65 of the IARC concerns drinking temperature. A wide, flat spout is supposedly more forgiving at the beginning: Sage UK advises the narrow one, and no source for the wide one was found. The jug may supposedly be filled up to the lower base of the spout: Sage UK and Lelit say half full. Foam supposedly separates from the milk within seconds: no measurement for this was found, and Lelit even lets it rest for half a minute. And many electric frothers supposedly deliberately make firm foam that can hardly be poured: as a blanket judgement this does not fit WMF, Severin and OutIn, which explicitly advertise latte art (for the Subminimal, the manufacturer sources we evaluated name two screens for different foam textures; we found no latte art promise there). A manufacturer statement that assigns a jug volume to a cup size was likewise not found, which is why our calculation appears above.