Watch enough talking-head footage and you will see the same flicker: the face brightens, dims, and settles again, sometimes mid-sentence. That is auto exposure doing its job. The meter reads the whole frame rather than your face, and the frame keeps changing — you lean in, a white shirt fills more of the picture, a cloud crosses the window behind you. Every change makes the camera re-decide how bright the image should be, and every re-decision shows on your skin. Autofocus does the same thing when you rock towards the lens and back, and auto white balance does it when your lighting is mixed. None of this is a fault. Automatic systems are built for scenes that change, and a person sitting still in a room is a scene that mostly does not — which is why the corrections are so visible when they come.
Two numbers set the exposure you are about to lock. Shutter speed is how long each frame gathers light. For speech video, keep it at roughly double the frame rate — 1/50 at 25 fps, 1/60 at 30 — so hands blur by the small, natural amount the eye expects. A faster shutter makes gestures stutter; a much slower one smears them. ISO is amplification. It brightens the picture by turning up the signal, and the noise comes up with it, first in the shadows and then everywhere. So the order of operations indoors is fixed: set the shutter to match your frame rate, then raise ISO only until the face reads properly. If the face is still too dark at an ISO you would be happy to watch, the answer is another lamp, not a bigger number.
Then take the decision away from the camera. Frame the shot, expose for the face — not the window, not the wall — and lock exposure. Lock white balance too if the light is mixed. Daylight from one side and a warm lamp from the other will drag auto white balance back and forth as you move, and skin is where viewers notice colour drifting. Once both are locked, the room can do what it likes: a cloud crossing the sun changes your background, and your face stays put. Locked settings also survive your own movement, which is most of what was fooling the meter in the first place — leaning towards the lens no longer makes you glow.
Talk2Camera puts these controls next to the prompter rather than in a separate camera app: ISO and shutter, exposure bias, exposure lock, manual focus and tap to focus, white-balance lock and stabilisation. If you switch cameras mid-recording, every control is re-checked against the incoming camera, so a value set on one lens cannot carry over out of range on another. Resolution and frame rate follow the same honest rule — 720p, 1080p and 4K at 24, 30 or 60 fps, and only the combinations your camera actually reports are offered, nothing greyed in that the device cannot do. For a seated talking head, 4K at 24 or 30 is usually the sensible ceiling; 60 fps buys a smoothness your delivery does not need, at twice the storage.
Leave everything on auto when the scene will not hold still. A walk through rooms with different light, an outdoor shoot under broken cloud, anything hand-held and moving — these are the scenes automatic systems were built for, and a locked exposure would be wrong within the first minute of any of them. Auto also wins when there is no time: a decent automatic take beats a beautifully exposed one you never recorded. The rule is not that manual is professional and auto is amateur. It is that a seated speaker under constant light is the one scene where the camera’s guesses are worse than your one decision — so make the decision there, and let the camera earn its keep everywhere else.