🔔 Update: Soliloquy V2 is here (*Bf16 still uploading...)

Soliloquy V2 is the next step forward for the model and directly addresses the main traits I identified in V1.

V1 established Soliloquy’s visual identity: dramatic photographic rendering, strong atmosphere, rich texture, and a distinct cinematic character. However, V1 also retained more of the underlying base model's original biases than I wanted, alongside a tendency toward overly aggressive contrast in some generations.

V2 was built specifically to address both.

It keeps the photographic identity that made V1 what it was, while pushing Soliloquy further toward its own visual character and giving the model a more controlled, coherent response.

What V2 does better

Compared to V1, Soliloquy V2 delivers:

• More natural tonal response
  Shadows and highlights are better controlled, with less tendency toward crushed blacks, overly harsh transitions, or pushed highlights.

• Reduced inheritance of base-model biases
  V2 has been refined to rely less heavily on some of the visual tendencies inherited from its underlying base, allowing Soliloquy’s own photographic character to come through more consistently.

• Improved detail hierarchy
  V2 is more selective about where detail belongs, producing cleaner and more photographic-looking images rather than pushing microcontrast everywhere equally.

• Better material and surface rendering
  Skin, fabric, metal, environmental surfaces, and fine textures feel more differentiated and naturally resolved.

• Stronger spatial depth and scene coherence
  Subjects, environments, motion, and background elements relate to each other more convincingly, especially in complex scenes.

• Cleaner light integration
  V2 handles dramatic lighting with more discipline, preserving mood and atmosphere while feeling less processed overall.

• Better action and motion readability
  Dynamic scenes retain their energy, but the intensity is better organized and easier to read.

• Better compatibility with Character LoRA's

In short, V2 does not reduce Soliloquy’s drama — it gives that drama more control, more realism, and more room to breathe.

Available Variants

Alongside the original (fp8) checkpoint, I've uploaded two additional variants:

• BF16 — a full-precision reconstruction of the model. Loads like any normal checkpoint using the standard "Load Diffusion Model" node (weight_dtype: default). No extra nodes required.

• INT8 ConvRot — a quantized variant (roughly half the file size) using INT8 weights with a Hadamard rotation applied to reduce quality loss versus plain INT8.

Overview

Soliloquy is a custom Krea 2 checkpoint created from a LoRA trained entirely on my own photography from my former studio, then permanently integrated into a carefully selected base model (see content notice further below).

It is not a conventional full-parameter finetune, but it is also more than a simple checkpoint merge.

Soliloquy is intended for cinematic portraiture, atmospheric environments, visual storytelling and surreal concepts that still feel as though they were captured through a real camera.

The Dataset

The LoRA used to create Soliloquy was trained exclusively on organic photographic data.

No synthetic or AI-generated images were used in the training dataset. Every image originated from my own photography, and every caption and tag was written by hand rather than generated through automated captioning.

Visual Character

Soliloquy tends toward:

• Cinematic and directional lighting

• Rich atmospheric depth

• Strong subject separation

• Natural skin and material texture

• Warm practical light against cooler environments

• Dramatic environmental compositions

• Photographic interpretations of surreal or impossible scenes

• A subtle analogue and DSLR-inspired character

The model is not limited to a single genre. It can move between portraiture, fantasy, fashion, landscapes, science fiction and surreal imagery while retaining a fairly consistent photographic eye.

V1 to V2

V1 should be understood as the foundation of Soliloquy’s public identity.

It introduced the model’s mood, contrast, atmosphere, texture, and photographic instinct, but it also retained a noticeable amount of the underlying base model's own visual biases.

V1 could sometimes lean too heavily into strong contrast, aggressive microdetail, and other inherited tendencies rather than allowing Soliloquy’s own photographic character to dominate the image.

V2 refines that foundation rather than replacing it.

The goal was not to turn Soliloquy into something fundamentally different, but to separate its own identity more clearly from the base beneath it.

V2 aims for:

• More control without losing mood

• More realism without flattening the image

• Less dependence on inherited base-model tendencies

• More coherence without sacrificing intensity

• Better material, lighting, and spatial relationships

• More polished scene construction while preserving Soliloquy’s visual voice

If V1 established the look, V2 is the version that lets it breathe.

Content Notice

Soliloquy is fully NSFW capable.

This is not a censored or SFW-only checkpoint, and V2 does not attempt to remove or suppress the broader generation capabilities of the underlying model.

There is, however, an important distinction between NSFW capability and unprompted NSFW behavior.

The underlying base model has a fairly aggressive NSFW bias and can sometimes introduce nudity even when it was not explicitly requested, particularly when prompts contain bodies, minimal clothing, bathing, lingerie, fantasy attire, or ambiguous wardrobe descriptions.

V1 inherited much of this behavior directly from the base.

V2 attempts to reduce that inherited tendency toward unsolicited nudity while retaining full NSFW capability when it is actually requested.

This does not mean that unprompted nudity has been eliminated entirely. Users seeking strictly safe-for-work generations should still describe clothing clearly and use appropriate negative prompting or workflow-level safeguard

Recommended Use

Soliloquy responds well to descriptive natural-language prompts, especially when the prompt includes:

• The intended light source

• Time of day

• Weather or atmospheric conditions

• Camera position and framing

• Materials and surface texture

• Emotional tone

• Foreground and background relationships

You generally do not need to overload prompts with long quality-tag strings. A clear scene, a strong visual intention, and a few carefully chosen photographic details tend to work best.

V2 in particular benefits from prompts that give it room to organize:

• Light

• Space

• Subject emphasis

• Environmental context

• Texture and materials

That is where the refinement over V1 tends to show most clearly

Closing Note

The name Soliloquy refers to the act of speaking one’s thoughts aloud.

This model is, in a sense, a continuation of that idea: old photographs, visual instincts and memories from a former studio translated into a new generative medium.

It is not an attempt to reproduce one fixed style.

It is an attempt to preserve a way of seeing.

Licensing & Disclaimer

• Original Model Creators: All credit goes to KREA.ai for the original research, architecture, and weights.

• License: This model is subject to the KREA 2 License Agreement. Please read and comply with the official license terms before using these weights: KREA 2 Licensing Terms.

• Official Distribution: This model is only officially available on Hugging Face and CivitAI. Unauthorised redistribution, rehosting, or modification of these model weights is strictly prohibited. If you encounter this model on any other platform, it has been uploaded without permission.

Basemodel used: Krea2TurboBadmilkmelancholy fp8 v1.0 by VINCE1968