A casting mold is selected first by the pour medium (resin, plaster, concrete, wax, low-melt metal) and only second by shape; the spec that drives reject rates is the dimensional budget between the master, the mold cavity, and the cured part [S1][S4].
For most non-metal pours the sprue or feed channel should be at least 3 to 5 cm (about 2 in) long to let the material flow and vent, and that minimum grows to roughly 5 to 8 cm for low-flowability media such as plaster, concrete, and wax [S3]. Sizing the cavity itself comes from a volume calculation: target part volume plus a 10 to 20 percent waste factor, divided by the cured material density, gives the grams of resin, plaster, or concrete to mix [S2].
Mold families and what each one actually does
The four working mold families in a casting shop are one-part (open-face), two-part, block (brushed-up shell), and brush-on (thin laminated skin over a backup); ArtMolds lists those four as the standard starter set, and ties each to a complexity range, not a brand [S4]. One-part molds suit flatbacks and pendants where only one face needs detail; two-part molds cover undercuts and full 3D sculpture; block molds are the default for simple geometric reproductions; brush-on molds are the only practical route when the master cannot be lifted out of a rigid shell [S4].
Material choice inside that family is driven by the pour medium, not by the part shape. Silicone captures the finest texture and tolerates resin, plaster, wax, and low-melt metals; latex is cheaper but tears faster and is limited to plaster and wax; urethane rubber bridges the gap for abrasive concrete and ceramic slurries; HDPE is a rigid, reusable option when the pour medium is resin and the run count is high [S1][S5].
Three numbers that decide every mold dimension
Three numbers drive the final cavity size: the part's bounding box, the cured-material shrinkage rate, and the draft angle on every vertical face. Plastics and resins typically shrink 1 to 3 percent by volume from pour to cure, which means a 100 mm dimension on the master becomes a 97 to 99 mm cavity before the elastomer's own thermal contraction is added [S2][S3]. Concrete and plaster shrink less, in the 0.5 to 1 percent band, so for architectural castings the cavity can be cut closer to the master [S2].
Draft is the second hidden spec: 1 to 3 degrees per side is the usual silicone-rubber window for resin, 5 to 7 degrees for plaster, and any vertical face below 1 degree will tear on demold no matter how soft the mold [S4]. The third number is wall thickness of the casting itself, which the pour-volume calculator has to read off the longest flow path: under about 5 mm wall, the pour freezes before it fills; over about 25 mm wall in a fast-cure resin, exotherm can crack the part [S2].
Pour-volume math and the role of the calculator
The shop-floor rule is to compute the cavity volume in cm³, multiply by the cured material's density in g/cm³, then add 10 to 20 percent for waste, mix losses in the cup, and the residue left on the mold walls [S2]. For a rectangular block mold this is straightforward: length × width × height × density × 1.15. For a complex two-part sculpture the standard workaround is to weigh the master in water (Archimedes) to get the displaced volume, then scale up by the inverse of the cured-material density [S2].
Once the pour mass is set, the sprue and vent geometry has to be sized to the same flow path. Prusa's beginner guide puts the minimum feed length at 3 to 5 cm for most resins and 5 to 8 cm for plaster, concrete, and wax, and the vent diameter is typically 1.5 to 2 times the sprue diameter to keep the air from pressurizing the cavity before the pour front arrives [S3]. These are the values a casting mold spec sheet has to call out before the cavity drawing is released.
Matching mold body to pour medium: a four-way comparison
Choosing the mold body is a four-way trade across pour medium, run count, detail resolution, and unit cost, and the same project can land on different families depending on which axis dominates. [S4]
Silicone rubber scores high on detail and works with resin, plaster, wax, and low-melt metal, but its per-part cost is the highest and its tear strength is the limiting factor on long runs; platinum-cure grades lift the temperature ceiling and the usable life [S1][S4]. Latex is the cheapest flexible option and is fine for one-off plaster and wax captures, but it cannot hold concrete or resin and degrades fast under UV [S1][S4]. Urethane rubber sits between the two: it tolerates abrasive concrete slurries, gives 20 to 50 demold cycles in a production setting, and captures detail almost at silicone level [S1]. HDPE is rigid, not flexible, so it is specified only for repeatable resin pours where the part has zero undercut and the run count justifies the tooling [S5].
The sand casting mold family is a separate branch used for molten metal: the binder (clay, chemical resin, or sodium silicate) replaces the elastomer, and the cavity is built up by pattern-and-flask rather than by brush-on or pour [S1]. The same shrinkage and draft rules still apply, but the upper temperature limit and the gas-evolution behavior of the binder become the binding constraints.
When a flexible silicone mold is the wrong choice
Silicone is the default and is often the wrong default: for long production runs of identical parts, injection or rotational molding in a metal mold base is cheaper per part once the run passes about 200 to 500 units, depending on part size [S1][S5]. For very large architectural castings, silicone rubber becomes impractical past roughly 50 kg of mixed pour because the mold flexes under the hydrostatic head and the part wall thickness drifts; urethane or fiber-reinforced plaster is the realistic body for those sizes [S1][S4].
HDPE rigid molds are the wrong pick the moment a part has any undercut or a re-entrant feature, because the part cannot be released without damaging the cavity; the same restriction applies to any one-part open-face mold used on a fully 3D sculpture [S4][S5]. Alginate is a fourth niche option that is fine for one-off body casting (life masks, hand casts) but cannot hold resin and degrades within days, so it sits outside any production workflow [S4].
Selection workflow from pour medium to shortlist
The shortlist logic, in order, is: pour medium, then run count, then detail, then unit cost. Pour medium eliminates about 70 percent of the candidate families on its own (silicone for resin and low-melt metal, urethane for abrasive concrete, plaster or sand for larger architectural and metal work) [S1][S4]. Run count then decides between a flexible mold (under a few hundred parts) and a rigid or metal tool (above that), with HDPE sitting in the middle for repeatable resin pours [S5].
Detail resolution and draft angle settle the final cavity drawing, and the pour-volume calculator fixes the mix batch size, the sprue length, and the vent diameter before any rubber is poured [S2][S3]. For buyers working alongside industrial kit on the same line, the spec discipline is the same one used to pick a mold temperature controller: sensor, setpoint, heater kW, and pump flow all have to be locked before the part drawing is released.
Trackable signals for the next six months
Three signals are worth watching through Q1 2026: (1) whether the major silicone suppliers publish revised pot-life and demold-cycle curves for their platinum-cure lines, since the working-time window drives batch sizing; (2) any update to the commonly cited 3 to 5 cm sprue minimum in published beginner references, because the value is older guidance and may be tightened for low-viscosity resins; (3) HDPE mold price movement, since HDPE has been the fastest-growing rigid-mold body for repeatable resin pours and any resin-grade shift will reset the break-even point against silicone [S3][S5].