First thing anyone who might click on an indie stream looks at is the avatar. If it looks like shit with garish colours and weird features then that's a red flag. It might signal to some that they're cheap and trendhopping so they don't give a shit about their model, it might signal that they're really new, it might signal that the streamer doesn't even have the most basic sense of taste or awareness about how they appear to people, which doesn't bode well for the stream. Sure, there might be cases where this isn't true, but most people won't bother to find out if they're already turned off.
After that they look at the title, but this has a different impact for each viewer. Uwu-speak will turn off some, titles worded implying there is already a solid culture might attract some while chasing off others. Clearly unintentional bad grammar or spelling errors are red flags.
After that is whether the voice fits the model. If a male voice comes out of a clearly female model most people just click away right there if they're ENtubers - babiniku culture is not strong. If a very deep alto comes out of a dainty fairy avatar many are also going to click away. Also, volume balancing and EQ plays a big role here in not making an unfamiliar voice not sound out of place next to music and game audio.
After all of that, people then start watching the content. One thing is that if the streamer speaks like two sentences a minute they're going to lose people fast, even if they're witty sentences. If they're already big it doesn't matter because a chat that knows them fills in the gaps - but if they're small the stream starts to look echoingly empty.
One mistake a lot of indies make is to greet every new face that pops into their chat by name and even address them with that. This makes prospective viewers uncomfortable because it gives off a sense of desperation, like you're trying to extract commitment from them by getting personal, akin to a sleazy car salesman who calls you by your first name every three sentences. It's also alien to other stream viewing experiences they've had. Answer questions and have conversations, but use names sparingly, if ever. The worst thing about automatically greeting new viewers is if the new viewer isn't a new face because they checked you out like a week ago and you just forgot. It makes you look artificial.
Also don't go too far on 'real life' stuff. The line is not easy to draw, but basically your viewer should never be getting the impression that they're just listening to a random twitch streamer without a camera. A vtuber avatar might be imagined to go to university classes or whatever and that's fine, but if a female vtuber says something like, "Yeah so I've done my nails up recently, I wish you could see it, I've spent a lot of time on it" - it's a topic that drives to the forefront of the viewer's mind the discrepancy between the roommate and the vtuber avatar. Needless to say sharing your real face is a surefire way of diluting and weakening your viewers' association with your 'brand' and kayfabe.
