Why marinade barely gets past the surface
Soak a block of firm tofu in soy sauce, garlic and ginger overnight, then cut it open: the center tastes like plain tofu, because after twelve hours the marinade traveled perhaps a millimeter past the surface.
The reason is water. Tofu is soy milk protein coagulated into a gel, and that gel already holds roughly 85 to 90 percent water before anyone does anything to it. A marinade is mostly water too, carrying salt, sugar and soy sauce compounds dissolved in it. Diffusion into any food follows Fick's second law: penetration depth scales with the square root of time, not with time itself, so soaking twice as long buys nowhere near twice the depth, and a food already saturated with water offers almost no concentration gradient to drive that diffusion inward in the first place. America's Test Kitchen ran into the same wall testing red wine against beef short ribs: even after 18 hours of soaking, the wine penetrated less than a millimeter. Tofu's dense, already water-saturated gel gives a marinade even less of a gradient to work with than muscle fiber does.
That reframes the whole job. Instead of trying to push flavor deeper into tofu, the more productive move is to pull water out, so the surface can concentrate whatever seasoning sits on it, dry out enough to brown, and let a coating actually grip.
What tofu actually is, and why the coagulant matters
Soy milk is mostly two proteins, glycinin and beta-conglycinin, floating in water with a slight negative charge that keeps them apart. Add a coagulant and its calcium or magnesium ions bridge those negative charges, pulling the proteins into a three-dimensional mesh that traps the surrounding water inside it. That mesh is the curd, and it is a gel rather than a solid for the same reason gelatin dessert is a gel: a protein scaffold holding mostly water in place.
Which coagulant does the bridging changes the result. A 2025 study comparing salt-based coagulants measured calcium-sulfate-set tofu as the softest of the group, describing it as fine and tender but lower in water retention, while magnesium-chloride-set tofu (the main salt in nigari) came out firmer, needing about 645 grams of compression force to deform versus visibly less for the calcium-sulfate curd, with a texture the researchers called looser and coarser at the gel level despite the added firmness. Blending the two coagulants produced the firmest tofu of all, which the same study attributed to calcium and magnesium ions binding at different sites on the soy protein at once. Glucono delta lactone sets slowest of the common coagulants and barely firms anything, which is why it is used for the jelly-soft silken tofu sold in shelf-stable boxes. One detail worth knowing when reading a label: calcium sulfate is literally a calcium salt, so a tofu set with it is contributing calcium to the finished curd in a way a magnesium-chloride-set tofu is not, independent of any texture difference.
Firmness grades describe how much of that trapped water was pressed back out during manufacturing, not a different soybean recipe. Silken tofu is set and sold in its own liquid with none drained off, soft and medium tofu have a little pressed out during the mold stage, and firm and extra firm are pressed under weight on the production line before packaging. Buying extra firm at the store is already a head start on the job a home cook would otherwise do with a towel and a plate.
Pressing removes water, not flavor
Pressing tofu adds nothing. What it does is force out some of the free water sitting in the curd's pores, water that would otherwise have to boil off in the pan before the surface can get hot enough to brown. Fifteen to thirty minutes under a steady, even weight is enough to visibly darken a folded towel underneath a block of firm tofu, and the payoff shows up later as a drier surface that browns faster and holds a coating instead of steaming it off.
Freezing does more than pressing ever could
Freeze a block of tofu solid and its water, still around 85 to 90 percent of its weight, forms ice crystals that expand as they form and physically tear apart the protein gel around them. Thaw it and drain it, and what is left is not the fine, uniform mesh it started as but a coarser, honeycombed structure with far more surface area and open channels running through it, the kind of change freeze-thaw research on soft and firm tofu has measured directly in the protein network. That ruptured structure drains water more completely than pressing alone, browns faster because less water is left to evaporate first, and takes up a sauce far more readily, not because sauce diffuses through protein any better than before, but because the tofu now behaves like an actual sponge instead of a sealed gel.
Japanese cooks found this by accident long before anyone had a freezer to blame. Buddhist monks at the temple on Mount Koya in the Kamakura period, sometime between 1185 and 1333, left tofu out overnight in the mountain cold, found it frozen solid the next morning, and cooked it anyway rather than waste it. The result, koya-dofu, became a staple of Buddhist vegetarian cooking and later shelf-stable enough to be issued as army rations during the Russo-Japanese War of 1904 to 1905. A home freezer at minus 18C does the same job in a few hours that a Kamakura winter did overnight.
The practical version: slice extra firm tofu into slabs, freeze it solid, thaw it fully, then press out the meltwater before cooking. The result has a meatier, chewier bite than fresh tofu and holds up better in a braise or a stir fry, which is why the technique shows up constantly in Chinese home cooking despite rarely making it into Western tofu advice.
Salt gets in where a marinade cannot
Salt is a small, fast-moving ion, and unlike the thick, sugar and soy sauce heavy liquid of a marinade, it can genuinely diffuse into the water already inside the tofu rather than sitting on top of it. A hot salted brine speeds this along further: the heat contracts the protein network slightly, squeezing out some of that internal water in a process called syneresis, which both lets a little more salt in and leaves the surface drier once the tofu is drained, so it browns faster in the pan afterward.
This is also why tofu cooked directly in a sauce or broth tastes more seasoned than any amount of cold soaking achieves. Cooking evaporates surface water continuously, concentrating the salt, sugar and umami compounds left behind against the tofu's surface as it happens, which is a different and far more effective mechanism than passive diffusion into a block that was never short on water to begin with.
How a starch coating actually gets crisp
Dust pressed or frozen tofu in cornstarch or potato starch and the starch granules absorb whatever surface moisture is left and swell, a process called gelatinization. As the coated tofu hits hot oil, that water boils off from the swollen granules, and the starch molecules that had spread apart lock back into a rigid, glassy, porous network as they dry, which is the structure that shatters into a crisp bite rather than bending. Cornstarch is the common choice over wheat flour partly because of its higher amylose content relative to potato starch, which builds a stiffer network once dry.
Underneath that coating is a temperature ceiling: as long as a surface is wet, its temperature cannot rise much past 100C, the boiling point of water, and the Maillard browning reaction that produces a deep golden crust and roasted flavor does not really get going until somewhere around 140 to 165C. A crowded pan makes this worse directly: too much tofu releases too much steam for it to escape, the whole pan's surface temperature stalls near 100C while that moisture cooks off, and the tofu steams in place rather than browning, no matter how long it stays over the heat.
Which tofu for which job
| Tofu type | Best use | Press it? | Failure mode if used wrong |
|---|---|---|---|
| Silken | Hiyayakko, soups, blending into sauces or desserts | No, it cannot survive pressing | Falls apart or shreds under a spatula in a stir fry |
| Soft or medium | Agedashi and other delicate simmered dishes | A light press at most, a few minutes under a towel | Turns rubbery if pressed hard, waterlogged if not drained at all |
| Firm | General stir frying and pan frying | Yes, 15 to 30 minutes under an even weight | Steams instead of browning if left wet |
| Extra firm | Crispy pan fried or baked tofu, skewers, scrambles | Yes, and freezing first improves it further | Still browns unevenly if the pan is crowded |
| Super firm, vacuum packed | Fastest weeknight crisping, no pressing step | No, it is already dense and low in water | Can dry out and turn chalky if overcooked, since little water is left to buffer it |
Getting the surface hot and dry enough
| Method | Temperature | Time | Why it works |
|---|---|---|---|
| Pan frying | Medium high, oil shimmering around 190C (375F) | 3 to 4 minutes per side | Hot, mostly dry oil lets surface water flash off fast enough to reach browning temperature before the tofu overcooks |
| Baking | 220C (425F) | 25 to 35 minutes, flipped once | Dry oven air slowly evaporates surface water across the whole piece, though it browns less deeply than a pan |
| Air frying | 200C (400F) | 15 to 20 minutes, shaken halfway | Fast moving air strips surface moisture almost as effectively as deep frying, with far less oil |
| Deep frying | 175 to 190C (350 to 375F) | 2 to 3 minutes | Full oil immersion brings every surface to browning temperature at once, giving the crispest, most even result |
| Boiling in salted water | A bare simmer, about 90C (195F) | 2 to 3 minutes before further cooking | Does not crisp anything, but firms the curd and lets a little salt diffuse in ahead of a hot pan |
None of this requires specialty gear, but a few things make the water-removal steps above easier to get right consistently.
-
A tofu press with even, adjustable pressure
A dedicated press applies steady, uniform force across the whole block, which pushes water out faster and more evenly than a hand-balanced plate and is less likely to crack a soft corner of the tofu.
-
A heavy skillet that holds its heat
Tofu releases a burst of steam the moment it hits the pan, and a pan with enough thermal mass keeps its surface temperature from collapsing under that moisture the way a thin, light pan can.
-
Reading a tofu label for coagulant and protein
The ingredient list tells you more about the block inside than the firmness label on the front does, since coagulant choice drives both texture and mineral content.
-
Water-packed versus shelf-stable tofu
The two are packaged for different jobs, and using the wrong one makes pressing or frying harder than it needs to be.
The honest complication: sometimes the wobble is the point
Everything above assumes crispy, dry-surfaced tofu is the goal, and across a lot of the food that made tofu popular in the first place, it is not. Sichuan mapo tofu is traditionally made with silken tofu, prized precisely because it barely holds together and takes on the sauce by simmering gently in it rather than by frying first. Japanese hiyayakko is a block of chilled silken tofu served straight from the water with grated ginger and scallion on top, and any firmer texture would defeat the entire dish. Agedashi tofu dusts soft tofu lightly in potato starch and fries it just long enough for a thin shell to form around a still-trembling, custardy interior, which is a controlled partial version of the crisping mechanism above rather than the full one.
The instinct that tofu must be pressed and crisped to be good is a preference that took hold in Western vegetarian and vegan cooking, not a rule the ingredient itself imposes. Judged against the dishes that use tofu on its own terms, un-pressed, wet, barely-there tofu is not a failure state at all.
This guide covers the technique and the food science of getting flavor and texture right in the pan. For where tofu, tempeh, seitan and TVP fit into a wider pantry, see the protein staples guide. For what the evidence actually says about protein adequacy on a plant-based diet, see the protein evidence page, and for the soy and hormones question specifically, see the soy and hormones evidence page.

