Character Design with AI - Real-Life Yoruichi Shihōin | Hyperreal AI Design Inspired by Bleach

Turning a stylized anime character into something that looks like it was shot on a real camera is one of the harder tests for an AI image generator. It is easy to get a generic fantasy portrait; it is much harder to keep the specific details that make a character recognizable while still convincing the eye that a real photographer was standing there. This piece walks through exactly that attempt, reimagining a fast-moving, golden-eyed anime swordfighter as a hyperreal photographic subject.

The published prompt is worth reading closely because it shows how much of the final image comes from precise, layered description rather than a single vague request. It specifies a tall, athletic build, golden eyes, dark skin, and long, flowing purple hair. It calls for a sleek, sleeveless outfit made of white and black robes. It names a mixed-race Asian identity for the subject. And it anchors the whole thing stylistically with two phrases that do a lot of work: "hyperrealist" and "in the style of National Geographic." That second phrase in particular is a common trick for pushing a generator away from illustration and toward something that reads as documentary photography, with natural light, candid posing, and photojournalistic texture.

What the oref parameter is doing

Alongside the descriptive prompt, the generation uses Midjourney's omni-reference feature, written as --oref followed by a reference image URL, with an omni-weight value set through --ow. This is the mechanism that lets a specific reference image influence the final render rather than leaving everything to text description alone. The omni-weight value of 155 sits on the lighter end of the scale, meaning the reference nudges the result without locking it rigidly to the source image. That is a deliberate middle ground: strong enough to guide the pose, coloring, or composition, but loose enough to let the detailed text prompt still shape most of the outcome. The aspect ratio is set to 9:16, a vertical format built for short-form video and mobile viewing rather than a traditional landscape portrait.

Why the character concept works as a design exercise

The source character is known for speed, discipline, and a striking visual signature built around purple hair, golden eyes, and flowing dark-and-white robes. Translating that into a photoreal image means the model has to hold onto a fairly specific and unusual combination of features, purple hair and golden eyes are not common in real photography, while still making the result look plausible as a photograph rather than a costume shot. The description of the finished piece calls out exactly this balance: tall, athletic, golden-eyed, and regal, rendered with what is described as candid realism. That combination of specificity in the prompt and restraint in the reference weight is what keeps the image from tipping into either a flat cosplay photo or an obviously synthetic fantasy render.

The spoken word layer

The video pairs the image with a short spoken word tribute rather than a straightforward voiceover explanation. The piece leans into motion-based language, describing the character sliding through near-darkness, flickering like something not quite fully present, moving with a rhythm that makes silence itself seem to keep time. It closes on the idea that even after the visible motion stops, something of that presence keeps going. Rather than narrating the technical process, the tribute treats the finished image as a character in motion that has simply been caught mid-step, which fits the source material's reputation for speed and flash-step movement.

What this demonstrates about prompt construction

Taken together, the prompt and the reference settings offer a small but useful case study for anyone building character portraits with Midjourney V7. Specific physical descriptors, build, eye color, skin tone, hair, and outfit, do more to anchor identity than a single style keyword could on its own. A style anchor like the National Geographic reference shifts the entire rendering pipeline toward photographic realism rather than illustration. And a moderate omni-weight value lets a reference image inform the render without overriding the text prompt entirely, which is a useful setting to know when a creator wants influence rather than replication.

Key takeaways

• The published prompt combines specific physical descriptors with a photographic style anchor to push the render toward hyperrealism rather than illustration.
• Midjourney's --oref parameter pulls elements from a reference image into the generation, with --ow controlling how strongly that reference is followed.
• An omni-weight of 155 is a relatively light setting, letting the reference guide the image without overriding the detailed text description.
• The 9:16 aspect ratio targets vertical, mobile-first video formats.
• The spoken word tribute reinforces the character's reputation for speed and near-invisible movement rather than explaining the technical process.

Who this is for

This piece is useful for anyone experimenting with Midjourney V7's reference tools, especially artists trying to render stylized or anime-inspired characters as photorealistic subjects. It is part of a broader run of AI character design and spoken word tributes from Humanitarians AI, which pairs original prompt engineering with short poetic pieces built around the finished image.