SuperAPI AI Directory

DENRAKEIW

DENRAKEIW is the personal portfolio website of AI engineer, AI production engineer, consultant, author, and artist Dennis Schöneberg. His work focuses on generative AI, creative advertising, and hybrid AI/CGI production, with the core being the construction of production pipelines for images, covering ComfyUI workflows, model optimization, LoRA training, and hybrid AI/CGI direction. He currently works part-time at Grabarz & Partner in Hamburg as an AI production engineer, while also providing AI engineering, consulting, writing, and artistic creation services to startups and studios worldwide as a freelancer. The website showcases multiple hybrid AI/CGI production cases, including Volkswagen Commercial Vehicles, the German Depression Aid Foundation, Porsche, Burger King, ProSiebenSat.1/Joyn, The Masked Singer, Ethel Vaughn, Hyundai, and E.ON, among other projects. In addition, he has released a standalone desktop AI image inpainting editor that supports selection, generation, color matching, and layer editing, and exposes more than 60 commands through MCP, which can be driven by AI agents.

Open website

About this tool

DENRAKEIW is the personal website of AI engineer, AI production engineer, consultant, author, and artist Dennis Schöneberg. His work sits at the intersection of generative AI, creative advertising, and hybrid AI/CGI production, with the goal of building the production pipelines behind images and delivering finished work for agencies, platforms, and artists, rather than just demonstrations.

In terms of purpose, Dennis Schöneberg mainly works in AI engineering and production: building ComfyUI workflows for creative advertising production, training LoRAs, optimizing models, and directing hybrid AI/CGI production. He currently works part-time at Grabarz & Partner (an award-winning agency headquartered in Hamburg that became part of the Omnicom Group in July 2023) as an AI production engineer, bringing generative AI into real creative advertising production. At the same time, he provides AI engineering, consulting, writing, and art services to startups and studios worldwide as a freelancer.

His main capabilities can be seen through his project cases. In the Volkswagen Commercial Vehicles project, the production pipeline built for Grabarz & Partner needed to keep the key visual usable when its long edge exceeded 15,000 pixels, so it used an upscale-first, then full-resolution manual retouching frame by frame approach to ensure that bonsai needles and cement bags remained intact at poster size. The German Depression Aid Foundation project used generative AI to let people perform all emotions while feeling none; apart from Enosch, who has real experience with depression, all people in the film were artificially generated, conveying the core message of "impossible to tell." The Porsche project is a series of hybrid AI/CGI productions, blending generative environments with photorealistic automotive CGI, with custom-trained models responsible for the fidelity of paint, metal, and light. The Burger King project produced two vertical ads; the dipping scenes were completed entirely by image-to-video generation, and the sauce replacement used WAN VACE video inpainting for shot-by-shot replacement. In the ProSiebenSat.1/Joyn project, the GNTM TopShot feature can transform a user's selfie into highlight model photos in 8 scenes within 5–10 seconds, backed by about 100 chained AI processing steps, including matting, perspective-aware composition, and skin and lighting retouching. The Masked Singer project generated introduction clips for costume characters using a custom ComfyUI pipeline. The Ethel Vaughn AI fashion project expanded a single shoot into 13 themes, 13 models, 13 locations, covering five continents, 100% clothing consistency, base resolution over 10k, editorial print size up to 6000×8000 pixels. The Hyundai Santa Fe global launch campaign let users generate dream destinations featuring the car via web and Instagram DM, with about 200,000 visits in the first month. The E.ON ugly sweater generator, meanwhile, built a standalone editor around its ComfyUI inpainting node.

He also released a desktop AI image inpainting editor that supports opening an image, selecting a region, entering a prompt, and generating. Results land as layers on the selection and remain editable: they can be color-matched to the surroundings, masked, partially erased, regenerated, stacked with filters and text layers, and all layers can be exported as PSD or OpenRaster. Selection tools include brush, rectangle, lasso, magic wand, click object (in-app SAM2), or text selection (such as "the handbag"). Generation uses "recipes" rather than node graphs, with a choice of model and where to run it (your own ComfyUI, local or RunPod, or an API provider with your own key). Each layer has a slider to blend the result into the image, and can color-match to the surrounding or underlying content. Filter layers run on the GPU, including film packs, text, 3D objects, and custom brushes. Every step is a command, so an agent can perform the same round trips. Currently supported on Windows, with an early build for Linux and macOS planned. A Rust-core tiling engine keeps 15,000-pixel documents usable; PNGs up to a gigapixel can be opened and saved in strips. The editor is free, licensed under GPL-3.0, currently an unsigned installer, with a signed version coming to the Microsoft Store in the future. The editor exposes more than 60 commands via MCP, and Claude Code, Claude Desktop, or any MCP client can drive it. The same commands are also provided to the built-in assistant: a chat bar beside the canvas that uses your own key (Anthropic, OpenAI, Google, or any OpenAI-compatible endpoint), can operate the editor itself, shows each call as a card, and asks before incurring costs. Every step is undoable, and an entire turn can be rolled back at once. The API adapters and assistant were written according to provider documentation and have not yet been tested against the actual services; local rendering through ComfyUI has been tested.

Applicable sceThe scope includes creative advertising production, AI/CGI hybrid production, generative advertising campaigns, model training and optimization, AI fashion imagery, AI music video creation, and more. His work has been exhibited in the QUBIT AI exhibition at FILE Festival 2024, and at the FIESP Cultural Center in São Paulo he presented four AI music videos created using SD 1.5 and Deforum, in which the synthesizer simultaneously serves as a camera control device, with music parameters connected directly to the generation process, driving the camera through 3D space, prompt weights, and model degrees of freedom; sound and image are not edited against each other, but come from the same signal.

In terms of honors, the Artificial Emotions project won the 2026 New York Film Festival LTX Competition Gold Tower, judged live by the 2026 NYF executive jury, led by Andrea Diquez (Global CEO of GUT), and the winning work stood out for innovative storytelling, creative execution, and pushing the boundaries of AI creativity. Other honors include winner of the ComfyUI workflow competition, first place on the vehicle creator leaderboard, third place on the tool creator leaderboard, Valentine's Day staff pick in the Lunar Love competition, ADC Germany Bronze, German Digital Award Bronze, Webby Awards Honorable Mention, and FILE Festival QUBIT AI exhibition.

Collaborators include Grabarz & Partner, Jung von Matt, BlackForestLabs, DaydaStudio, ossa.ai, gleem.ai, rendermind.ai, kartel.ai, Lighthouseacademy.ai, The Customization Group, Mintano, Ethel Vaughn, and others. Dennis Schöneberg also serves as a lecturer at Lighthouse Academy.

In terms of limitations, the website does not provide specific contact information, service pricing, complete product documentation, or detailed tutorials. The desktop editor currently has an official version only for Windows, Linux is an early build, and macOS is planned; the installer is unsigned. The API adapters and assistant have not yet been tested against actual services. More information is available through his Civitai and GitHub pages (DenRakEiw).