Great info, saved!

> I’m struggling to get that truly lifelike quality for character portraits. Most of what I produce still has that slightly 'plastic' or uncanny valley look.

So basically, you want that DSLR vibe without the headache, right? Well, I mean, if you're worried about the learning curve of complex tools, you might wanna look at Leonardo.ai. Honestly, it's pretty beginner-friendly compared to others. You can use their "PhotoReal" mode which is basically a toggle that fixes that weird plastic sheen. It costs about $10/month for the basic plan, which is decent value. Just be careful with high-res stuff cuz it can eat credits fast, you know? Good luck!!

+1

Seconding the recommendation above for Flux.1 [pro]! Honestly, it's been a game-changer for my local projects. Since you mentioned wanting that DSLR look, I've found it depends on where you're based and the 'vibe' of your local lighting. Like, if you're aiming for that overcast, moody UK photography style vs bright California sun, you gotta tweak the prompts for atmospheric haze.

Here's a quick comparison of what I've tried:

1. Flux.1 [pro] vs Midjourney v6: Flux is way more 'honest' with skin texture. Midjourney is beautiful but always feels a bit like a magazine cover, you know? Flux actually gives me those tiny pores and stray hairs without it looking like a filter.
2. Leonardo.ai PhotoReal: This is probably the most cost-effective for a beginner. The daily free credits are decent, but the paid tier is worth it for the consistency. It's way easier to handle than Stable Diffusion if you don't wanna spend hours on settings.

Basically, if you want that 50mm f/1.8 shallow depth of field, try adding 'chromatic aberration' or 'film grain' to your prompts to kill that plastic sheen. It works SO well. Are you planning on printing these or just using them for digital stuff? That might change which subscription makes the most sense long-term. gl! 👍

Honestly, I get exactly what you mean about that plastic look. Over the years I've tried many different tools, and for your situation, I'd suggest checking out Flux.1 [dev] or the Flux.1 [pro] model. I'm still kinda new to the technical side of Flux, but basically it handles skin texture and lighting way better than Midjourney v6 without that weird artificial sheen. Ngl, it's a total game changer for realism.

If you want a more "out of the box" experience with professional headshot quality, I recommend Leonardo.ai with their PhotoReal setting turned on. It gives you great control over the 50mm lens look you're after. Just remember to add "high-resolution skin pores" and "natural eye reflections" to your prompt to kill that uncanny valley vibe. It's way easier than mastering Stable Diffusion but gives you that DSLR feel you want. gl!

Curious about one thing: are you needing full-body shots too or just close-ups? tbh I've found Leonardo.ai is way easier for faces than SD, but I gotta know the framing first! TL;DR: Leonardo's PhotoReal mode is goated for skin texture.

Curious about one thing: studio or outdoor lighting? Over the years, I've tried many and some brands just struggle with shadows. My current setup actually NAILS that DSLR vibe tho.

Yeah, I totally agree with the folks suggesting the Flux architecture—it’s current SOTA for subsurface scattering and realistic skin pore reproduction. If you’re looking for a budget-friendly way to get those pro results without a massive monthly commitment, check out Fal.ai or Together AI. They offer an API-based 'pay-as-you-go' model that’s way cheaper for testing than a $20 or $30 a month subscription. You basically pay a fraction of a cent per inference, which is great for the wallet. From a technical side, that 'plastic' look usually stems from an over-reliance on high-denoising strengths or poor lighting math in the base model. To get that genuine 50mm DSLR feel, try specific technical tokens like 'f/1.8 aperture,' 'ISO 200,' and '8k UHD textures.' Pro tip from the community: look into using specific 'Realism' or 'Film Grain' LoRAs on these platforms. They add that high-frequency noise and slight lens distortion that makes a portrait look like it actually passed through glass rather than being rendered. Tbh, it’s much more cost-effective to fine-tune your parameters on an API where you only pay for what you actually generate. Reallyyy helps when you're still in that trial-and-error phase!

Ok so, training data and lighting math are everything!! Quick question: are these for commercial use?? I finally fixed that artificial sheen in my own work recently, but you gotta be cautious with legalities!!