The Intersil HIP4080A controller chip doesn't appear to need the normal "freewheeling diodes" that are sometimes seen in motor driving circuits. The diodes in parallel with the MOSFET's in Figure 33, Page 14, are integral to the N-Channel MOSFET itself and not a separate component.
If, for whatever reason you decide that you need to use freewheeling diodes, you would put them in parallel with the ones integrated with the MOSFET's. I would recommend diodes with a reverse voltage of at least double your motor voltage and 1x, and preferably 2x of the stall current of the motor. Since your motor is 12 volts @ 25 amps, I would recommend diodes of at least 24 volts and 50 amps. These are not small diodes. The diodes in the datasheet you linked to are not even close to handling 50 amps.
In Figure 33, there are two more diodes, CR1 and CR2, but the datasheet doesn't say how to spec these. The diodes are located between +12v and AHB/BHB pins. On page 4 the datasheet says that the absolute max voltage on those pins is 80v+VDD, and since VDD is 12v then the max voltage is 92v. Therefore, without going through all the little details you'll need a diode that is spec'd for a reverse voltage of at least 92 volts.
(Note: I'm doing a very rough analysis to come up with 92 volts, and I'm being very conservative too. If you get a diode that goes to 92 volts then it'll work. It very well might work with a diode rated at only 25 volts, but it's impossible to tell just from this datasheet. To be sure, you'd either have to get more info from the chip manufacturer, simulate it, or build it and measure the thing.)
It's also really hard to tell from the datasheet what the current ratings on CR1&2 need to be. It is mostly going to be determined by the values of C3 and C4-- values that they don't give. Here again you'll either need more info from Intersil or you'll have to build it, measure it, and then change the diodes accordingly. As a rough starting point I would go with diodes in the 1 or 2 amp range, but keep in mind that I could be as much as 10x high or 10x low.
Plain H bridges can be used to control large steppers, provided that they have the current/thermal capacity. But it's not efficient to do so.
The problem with a stepper motor is that the windings have lots of reactive impedance, and a motor with fine steps, rotating at or above a moderate speed, will be trying to switch the current flowing through that inductance very quickly. Doing this requires a quite high voltage - eventually many times the voltage necessary to push rated current through a stationary coil which shows only resistive impedance.
The designer of a simple driver has a choice: they can size the voltage for the stationary case, and lose torque (and soon miss steps) as the step rate increases. Or they can size the voltage to overcome the inductance of the high speed case, and overdrive (and overheat) the motor when it is not turning.
An early solution was to use a very high voltage, and huge power resistors in series with the coils - in effect reducing the ratio between the total impedance in stationary and rotating cases. This was actually done on some early CNC conversions of full size bridgeport milling machines, but effectively means there's a resistive heater strapped to the back of the cabinet.
The modern, efficient solution is a chopping current drive. This is effectively an additional circuit which rapidly enables/disables an H bridge. When a step occurs, the winding is energized at a high voltage. A comparator then monitors the rise of current though the winding inductance over time (typically by sampling the voltage on a high power fractional-ohm sense resistor). When the current has risen to a set point level, the driver is disabled and the current falls. It's then re-enabled and the cycle repeats - as long as a given winding is commanded to be energized, it will be "chopped" on and off to achieve the specified current.
Ultimately a chopping drive is an H bridge - but one with an extra current regulator inserted between the step generator and the control signals to the FET's comprising the bridge.
NEMA23 is about at the dividing point for H bridge construction - anything much larger and you want an assembly of discrete power FET's, while for limited applications at that size and lower (desktop 3d printers, etc), you can probably use an integrated circuit bridge or complete driver circuit with chopper included.
Best Answer
Your diodes V rating should be high enough to withstand your generators/any other stuff you added to the system. For you, they should be at least 30VDC rated. Of course, for practical purposes, never go that low. Pick something with 50VDC~100VDC. Also, diode amper rating must be selected using the same principles. If you will draw 1 amps, get diode rated for at least 1 amps. I'd get 3 amps.
So something like 100 VDC, 3 Amp rated Schottky Diodes are okay for you.
Something powerful enough as described earlier. Remember, you draw current. If you plan on drawing 100 mW from a 1000 W source, you can! So the rectifier power specifications demand on your load, not the source. (Of course it should withstand source's voltage, though)
There is a graph, of how many amps can kill you per time. It is something like 10 mA ~a couple seconds to 100 mA ~a couple microseconds. So dangerous limit starts at 10 mA but thats not enough. Why? What causes electrons to penetrate human body? Voltage! Dangerous voltages (the ones that hurt you) start from 50V. Any source rated above 50V and has amp rating more than 10mA has a chance of killing you depending on point of contact. Example: Getting electrocuted from your left hand-feet is worse than right hand-feet because your heart (generally) is on the left side. Amperage ratings differ from people to people etc. so this is not a simple question to answer. But like I said, let's assume that any source rated above 50V and 10 mA has enough power to kill a grown man.
In your case, your source is 30V which is not enough to penetrate your body but I would be cautious. Use plastic gloves at least. You can also work on an insulated ground elevated above from real ground as used in labs.