<h2 id="description">Description</h2>DepthStage is a hybrid realism-to-manga style LoRA designed to place photorealistic subjects naturally inside illustrated manga environments.The model was developed through repeated testing across people, animals, indoor scenes, villages, mountains, coastal environments, fantasy landscapes, safari scenes and everyday working environments. The goal was not simply to mix two visual styles. The goal was to control where the realism appears and where the manga treatment appears.A realistic person can stand naturally inside a hand-drawn manga village. A photorealistic elephant can walk through an illustrated African savanna. A realistic worker, traveller, horse or foreground object can retain convincing materials, anatomy, texture and photographic lighting while the landscape behind it develops into a colourful hand-drawn world.The most important feature of DepthStage is depth separation. The foreground subject can remain realistic while the middle ground and background progressively adopt the manga treatment. Perspective, lighting, shadows and colour relationships help the two visual languages belong to the same scene instead of looking like separate images pasted together.DepthStage can also produce interesting intermediate results where some secondary subjects become illustrated while the principal subject remains realistic. This makes the LoRA useful for experimentation rather than locking every generation into one fixed appearance.<h3 id="recommended-prompting">Recommended prompting</h3>For the strongest hybrid effect, clearly tell the model what must remain realistic and what must become manga. Do not simply write “realistic manga,” because this can cause the two styles to merge everywhere.A useful structure is:[main subject] + photorealistic treatment + exact position/action + [environment] + manga treatment + depth relationship + shared lightingFor example:Positive prompt: <code>DepthStage, photorealistic elderly fisherman in the foreground repairing a fishing net, realistic face, hands, clothing, rope and natural skin texture, Mediterranean harbour village behind him rendered as detailed hand-drawn manga illustration, colourful houses, boats, stone steps, blue sea and large illustrated clouds, foreground fisherman remaining distinctly photographic while the middle ground and distant environment become progressively more manga-stylized, coherent perspective, matching sunlight and shadows, natural integration, strong depth, cinematic composition, 1024x1024</code>For animals:<code>DepthStage, large photorealistic African elephant walking in the immediate foreground, anatomically accurate body, realistic wrinkled skin, tusks, dust and natural sunlight, expansive African safari landscape behind the elephant rendered as colourful hand-drawn manga illustration, illustrated acacia trees, mountains, clouds, zebras, giraffes and antelope in the middle and far distance, realistic elephant clearly separated from the illustrated environment, shared golden-hour lighting, coherent shadows, atmospheric depth, 1024x1024</code>For people:<code>DepthStage, photorealistic adult traveller standing prominently in the foreground, natural face, realistic skin, hair, clothing and backpack, looking toward camera, vast mountain village behind the person rendered entirely as colourful hand-drawn manga scenery, illustrated stone houses, winding paths, mountains and large expressive clouds, realistic foreground person against manga middle ground and background, coherent perspective and matching daylight, seamless depth transition, 1024x1024</code>For indoor scenes:<code>DepthStage, photorealistic elderly woman preparing vegetable soup in the foreground, realistic aged face, hands, clothing, vegetables, steam and cooking pot, rustic country kitchen surrounding her rendered as detailed hand-drawn manga illustration, illustrated wooden shelves, cookware, fireplace, cupboards and window, warm interior lighting shared naturally between realistic woman and illustrated environment, strong foreground-background separation, 1024x1024</code><h3 id="negative-prompt">Negative prompt</h3>Using the negative prompt is strongly recommended with this model. It helps prevent the realistic subject from being converted completely into manga and reduces the chance of the background becoming photorealistic.Start with:<code>cartoon foreground subject, anime foreground subject, manga foreground subject, illustrated main subject, cel-shaded main subject, flat-shaded main subject, plastic skin, doll face, 3D character, CGI character, unrealistic anatomy, malformed hands, extra fingers, extra limbs, duplicated subject, distorted face, crossed eyes, blurry face, low-detail face, photorealistic background, photographic manga scenery, realistic background landscape, pasted cutout appearance, collage effect, split-screen composition, hard dividing line between styles, inconsistent perspective, mismatched lighting, conflicting shadows, floating objects, distorted architecture, warped buildings, duplicate animals, malformed animals, text, watermark, logo</code>The negative prompt should be adjusted according to the generation. For example, when generating a realistic elephant against a manga safari, emphasize:<code>cartoon elephant, manga elephant, anime elephant, illustrated elephant, deformed elephant, extra legs, duplicate elephant</code>When generating a realistic human:<code>anime face, manga face, cartoon face, illustrated skin, doll skin, malformed hands, extra fingers, distorted eyes</code><h3 id="important-technique">Important technique</h3>DepthStage responds best when the prompt establishes the visual hierarchy clearly:REAL SUBJECT → POSITION/ACTION → MANGA ENVIRONMENT → DEPTH → SHARED LIGHTINGDo not overload the prompt with many competing style words. Decide first which element is the realistic anchor.For stronger realism, use phrases such as:<code>photorealistic, natural skin texture, realistic material detail, photographic lighting, anatomically accurate, natural imperfections, physically realistic</code>For a stronger manga environment, use:<code>hand-drawn manga background, illustrated scenery, ink-defined architecture, cel-painted colour, simplified hand-painted shading, expressive illustrated clouds, traditional manga environment design</code><h3 id="controlling-the-strength">Controlling the strength</h3>If too much of the image becomes manga, strengthen the realistic subject:<code>foreground subject remains strictly photorealistic, realistic photographic face and body, do not illustrate the main subject</code>If too much becomes realistic, strengthen the environment:<code>entire background and middle ground rendered as hand-drawn manga illustration, illustrated environment only, clearly visible ink lines and cel-painted colour</code>If the two styles look pasted together, add:<code>same directional light, matching colour temperature, coherent cast shadows, unified atmospheric perspective, realistic depth relationship, naturally integrated into one scene</code><h3 id="what-depthstage-is-good-for">What DepthStage is good for</h3>DepthStage works particularly well for travellers, portraits, workers, villagers, horses, elephants and other animals, mountain settlements, Mediterranean villages, coastal scenes, safari landscapes, forests, fantasy environments and indoor scenes.It can also be reversed or deliberately pushed into unusual combinations. Experiment with realistic foregrounds, illustrated secondary characters, realistic animals against manga landscapes, or different levels of manga influence through foreground, middle ground and background.The model is designed to be experimented with. Changing the LoRA strength, prompt order, negative prompt and the explicit division between REALISTIC SUBJECT and MANGA ENVIRONMENT can produce substantially different results.Trigger word: <code>DepthStage</code>Recommended starting concept: <code>DepthStage, photorealistic [SUBJECT] in the foreground, [ACTION], detailed hand-drawn manga [ENVIRONMENT] in the middle ground and background, realistic subject remaining photographic, manga environment clearly illustrated, coherent perspective, shared natural lighting, strong depth separation, 1024x1024</code>The central idea behind the model is simple:<div data-youtube-video><iframe width="640" height="480" allowfullscreen="true" autoplay="false" disablekbcontrols="false" enableiframeapi="false" endtime="0" ivloadpolicy="0" loop="false" modestbranding="false" origin playlist src="https://www.youtube.com/embed/11JL0ti5bB4?si=-s3nQts8ZA4o0l3C" start="0"></iframe></div>Keep reality where you want reality. Put manga where you want manga. Use depth, perspective and light to bring them together.

Files with SHA256: 470fd5441c34b79617585f31d339322b0e1a4e388b67b136189ab5ed24c68d29