I think the "what makes sense for the story" is a cop out. That's always going to be the case. To me there are two issues. When can you see a humanoid sized figure, and when can you identify what kind of creature it is.
So there are many, many variables at play. If there's nothing to gauge distance, a giant might appear to be a normal size human. A halfling might appear to be a human much further away than they actually are. What kind of terrain are we talking about? What are the lighting and atmospheric conditions?
Under ideal situation you could see figures more than a mile away (3 miles or further if you have elevation). Having lived in the country I'd say the odds of picking out a lone individual at that range unless it was a dark figure on a snowy plain is nil. I'd personally set the max distance for normal vision somewhere around a mile and a half, but that's just my guesstimate and will vary widely. In many cases it will be less because the figure will blend into the background.
But what if we're talking about recognizing that the humanoid is an orc. After a quick google I found an article
here that states "At about 150 feet, accurate face identification for people with normal vision drops to zero."
We aren't trying to identify a specific individual, just whether it's an orc or a person approaching. But unless the orcs are carrying a giant "We're orcs" banner I don't think it's going to increase that much. Let's say you could potentially tell the difference between an orc and a human at 200 feet.
So I'd adjust it from there. At 200 feet, the DC is 25, dropping as you deem appropriate. Maybe 20 at 150 feet, 15 at 125, 10 at 100, automatic after that.
Which of course will all be modified by whether the orcs are carrying that giant banner, have made any effort to disguise themselves with a hooded cloak, light conditions and so on.