So I have had a good think about possible biology for plausible (non-magical) bird humans. I am not going to go into much detail of how they evolve, because the proposed model of being like a stereotypical angel, with separate arms and wings, would be extremely implausible to evolve when all other vertebrates operate the same 4 limbs principle. A true flying humanoid is more likely to be like a bat, and not having access to the arms for operating weapons would drastically reduce their combat potential, with dive bombing and blind bombing using simple drops from handheld munitions being the only practical aerial attack methods ( I doubt they would be able to use belts/pouches whilst flying). In a bat-like scenario, they would be limited to the (still very useful) intelligence and communication roles.
First of all- size and weight.
The heaviest flying (and arguably largest) known bird ever is
this beast, Argentavis magnificens. It is estimated to have weighed about 70kg, a height of 1-2m and have a wingspan of 5-6m. This conveniently puts it at about the size and weight of a skinny human with wings instead of arms. So as a very quick proof-of-concept, flying humans could work on a biological level.
We can expect a humanoid that is capable of reasonable function on the ground (so walking and using arms as a normal human can) will have a much greater weight burden than a bird, which is focussed on flying. Therefore our flying humans have several options:
An even greater wingspan than above (this has increasingly diminishing returns)
Some reduced function over normal humans, the easiest would be very underdeveloped legs which would result in much poorer walking ability
Reduced size of the human attached to the wings, to bring the weight down.
The info we have been given about these birdpeople is that they are similar in size to normal humans, and have similar capabilities to humans, but with the addition of flying. I think it is reasonable to expect a degree of reduced function in walking and running endurance, due to lack of use compared to a conditioned human, and that could drop the weight by 10kg or so at most.
Lets assume a slim human of roughly average height, about 50kg (7st. 12Ib) and 1.6m (5' 3"). This is without wings. The size of wings that would be needed to support this is pretty big. Birds apparently have a wing-loading between 1-20kg/m2, and the theoretical upper limit to maintain avian flight is 25kg/m2. The Argentavis bird above had a wing loading of 8.6kg/m2. At a wing loading of 20kg/m2, the wing area would have to be 2.3m2 just to support the body, not including the wings themselves. However, this seems to be for ungainly birds like waterfowl and waders, which don't need good aerodynamics. Our flying humans probably have to be fairly manoeuvrable, so lets use the wing loading for Argentavis, as it was a bird of prey and may have hunted.
Working out the weight of the wings themselves has been challenging. However, I found
this source, which states that the wings of a bald eagle weigh about 0.9kg for a 2.3m wingspan, and I've
discovered that the similarly sized and closely related white-tailed eagle has a wing area of about 0.9m2. This means about 1kg per m2 of wing. From this, we can work out a very rough estimate of how much wing is needed to carry our birdperson with a wing loading of 8.6kg/m2. This gives 5.8m2 for our human weight, and a wing area of about 6.6m2 when including the weight of the wings. Using the size of
Argentavis' wingspan compared to wing area, we can work out a wingspan of roughly 7.6m.
This gives the total size of our bird people as about 55kg, with a wingspan of 7.6m. This is not hugely bigger than Argentavis, so we can assume our bird people are reasonably manouevrable, but would have great difficulty in taking off. This is their most vulnerable time. They also would be far less agile than falcons and eagles and hawks, and would only be able to sustain rapid flight for brief periods (flapping the huge wings). This is likely also an underestimate- I've not taken into account the differences in musculature across the shoulders and chest needed to operate the wings, and this would likely add a good few kilos more, and correspondingly more wing space (
apparently, flapping muscles require ~17% of a bird's bodyweight). In addition, the wings themselves may need to be a bit heavier near to the torso to support their own weights, so my wing weight estimates are likely too low. I may come back and redo my calculations to take into account a beefy chest for wing flapping.
Speed needed to maintain flight.
Argentavis apparently needs at least a speed of 11m/s (24.6mph, 39.6kph) to sustain flight. I would suspect a similar speed would be needed for our flying human, as we have given it the same wing loading.
The cruising speed for large birds is around this speed too, so basically our bird people would likely stay at about 25-30mph the majority of the time. Interestingly,
the maximum gliding speed increases with the size of the bird, so the bird folk could glide as fast as 20mph with little effort. With favourable conditions (slope and thermal soaring) they could reach speeds of 40mph, but this would only be sustainable as long as these conditions persist. Slope soaring requires mountains and hills, so mountainous regions would increase their mobility drastically. The upper limit of speed for birds in a dive can vary considerably, and large birds like eagles can reach 200mph, but I doubt our flying humans would be able to tolerate these speeds with their excessively large wingspan, and would struggle to pull out of such a dive without damaging something or doing it very gradually (so high off the ground). In addition, after a dive, they would struggle to regain altitude without a thermal, meaning they would be vulnerable to ground fire after they stuka-dive enemies. Related to this, the maximum altitude of such large birds is often around 5000m, especially if they can hitch lifts on thermals. Bird people would not be able to reach those heights easily, would be able to see very little, have little ability to hit even an army and would take awhile to descend. I think these heights would only be used for long-range scouting and observation. An operating height of about 500m seems much more practical for combat purposes, and if they operated their supply bases from cliff-tops, they would be able to land and take off easily. Flat plains areas would be the regions they expend most energy to operate (gaining altitude) and take the longest to get into position. In fact, it may be impossible for them to take off from flat terrain without a suitable headwind to generate lift. These are the best regions if you want to defend against them or attack. Mountains would be a nightmare to combat them. Attacking on a windless day would also make it more difficult for them.
Next up- power output and energy consumption.
So, again using Argentavis as our best available comparison, this beast requires 3.5 times more energy than it could sustainably output. The flying humans would need even more energy to maintain powered flight, and would likely be even more underpowered for flapping due to it's shared biolgy with arms. So from this, it is safe to assume our birdpeople are gliders, that can only sustain flapping for a brief period (such as takeoff) before suffering fatigue. This makes their combat potential very limited in endurance, unless they simply glide over. There is going to be no agile dives and regaining altitude quickly, at most short dives and dropping payload, before leaving the area to rearm and report back observations. Because of this lack of available power, their flight paths are also going to be considerably more predictable, because they can only sustain erratic course corrections briefly. Therefore firearms could be reasonably effective shooting on-mass.
The power estimated for Argentavis to maintain flight is 600w, but is negligible when gliding or soaring. Our bird people have larger wings and a less efficient aerodynamic weight, although are slightly lighter, so their necessary sustained power could well be considerably higher. I don't have the measurements or algorithms needed to calculate this, but I would not be surprised if it was double. Lets say 1,000w. This is 1,000J per second of active flying, so take-off and combat manouevring is going to be hugely energy intensive- 1 minute of active flying is 60kJ. A normal human requires at least 7,500kJ per day, according to the World Health Organisation, so that is the energy level for basic human metabolism plus some daily activity. Combat flying is likely to add several thousand more kilojoules of energy to that total, maybe even doubling it, if they run a lot of sorties. This is possible to eat, humans can consume 15,000kJ of energy per day, and have to in conditions like cystic fibrosis. This is going to require essentially 50% more rations for actively flying troops per day, and they are going to exhaust themselves pretty quickly in repeated attacks with their low sustainable power threshold. Launching from a clifftop would reduce the energy costs considerably and aid logistics. Simply scouting and remaining at altitude could be far more efficient than walking though, so using the flyers for reconnaisance would be logistically easier.
When laden with a combat load, taking off on the flat would be impossible without assistance (catapult maybe?) and gaining altitude would require thermals and wind with the poor sustainable power. Combat flying would also be drastically affected until munitions are offloaded. Armour at best would be thin cloth (this could prove fatal with rain), but that may be enough to protect against light arrows at the top of their arcs at higher altitudes.
Combat
Active flying largely uses the pectoralis muscle on the anterior chest wall. This muscle also acts on human arms, but is not that involved in throwing or drawing a bow. In gliding flight, it could be used to stabilise the shoulder whilst light weapons are thrown or drawn, but it would be impossible to utilise the back muscles for throwing/drawing a bow without destabilising the flight. It may be possible to accept a temporay drop and then recover into stable flight, but I think in practice gravity will be doing most of the work for projectiles. A dropped soliferrum would still be effective, although a dropped pebble is going to be fairly unimpressive. Medium sized rocks are probably the minimum for useful damage, other than caltrops. The limited munitions due to weight is also an issue, along side balance and storage. Flying with a bag would be disruptive to aerodynamics, so I wouldn't be surprised if such flyers were limited to just an item in each hand, and maybe some small gear on belts. In addition, incendiaries would be pretty much impossible to light on the wing, and fuses could only be lit prior to flying if the supply base is near to the combat. Thankfully, I think these are pretty limited in scope, unless the bird folk develop some kind of working percussion fuse in a suitably lightweight device. If they do, making the fuses unreliable could be a good way to balance this.
Summary
The bird folk could fly, and could eat enough to do it, but their combat capabilities would be limited without nearby resupply, they would tire easily in such conditions, would be very vulnerable without suitable terrain to take off, and their poor ability to dodge and enormous wingspan would make them huge targets if they come within range of the ground folks' weaponry. Huge vulnerable targets, seeing as it wouldn't take much damage to a wing to make it either lame or unable to sustain sufficient lift. Any such injury whilst flying would likely result in a greater proportion of casualties than ground troops too, so the risks are greater. Scouting, communication, observation and spying seem like by far the best options for employing flying troops, as well as ambushes in favourable regions. They would rule the mountains, but fear the plains.
