I build nodes as lenses: systems for directing models, preserving continuity and turning generative video into an authored production process.
Most of these systems began as private solutions for my own films, tests and impossible timelines. They are not frozen products. They are working instruments — rewritten whenever a production asks a better question.
Some live in the open IAMCCS node ecosystem; others are available as subscriber editions or perpetual licences. What joins them is the same idea: the machine should move closer to the filmmaker, not the filmmaker closer to the machine.
A local-first image studio built around your own ComfyUI: generation, reference-led creation, multi-region inpainting, outpainting, layers, composition, finishing and assistant-driven control. A larger creative surface, without handing your images or production logic to a black box.
Not a face generator, but a production room for repeatable identities: casting, visual direction, character bibles, controlled variations, datasets, campaigns and persistent projects. The point is not to manufacture a person. It is to direct a visual identity coherently across time.
Direct the timeline instead of gambling on a single prompt. Global direction, local shot intent, positioned guides, timing, audio, continuity and technical behaviour remain visible in one authored system. The board becomes the source of truth; H3 becomes the camera department.
The last frame is not the whole memory of a shot. IAMCCS long-video tools work around temporal continuity, latent continuation, overlap, first/last-frame control, positioned guides and disk-backed extension — trying to preserve not just appearance, but the residue of movement, rhythm and audiovisual state.
The edit should not begin after generation. Shotboard planning, AudioBoard, multi-timeline routing, dialogue, master audio, take management and the video editor push editorial thinking upstream — into the moment in which the images are still being born.
Generation is not the final image. CinePostEFX is built for the last passage through the darkroom: grain, sensor character, lens behaviour, tape, broadcast texture and the small imperfections that let an image stop looking generated and begin to carry atmosphere.
goyAIcanvas grows beyond the still frame.
Pianosequenza is the next room: a video-production system conceived around shots, continuity, direction, sound and editorial time. Not an endless generate button — a place where an AI sequence can begin to behave like a production.
These tools keep changing because the projects keep changing. I stay close to the people who use them, together with my team, through continuous builds, documentation, direct assistance and production feedback. A subscription is not simply access to a file: it helps keep the instruments alive while they are being tested against real creative work.
Tools, licences, installation and support: goya@carminecristalloscalzi.com · General information, film/art collaborations and other enquiries: info@carminecristalloscalzi.com
A first walk through the local image generator and editor: from generation to Imagine, editing, composition and the production logic behind the workspace.
Subgraph-aware drawing, text, layers, screenshot post-its, selection, transform and rotate tools directly on the ComfyUI canvas.
A visual prompt selector for Qwen Image Edit, Flux-based image editing and dataset-generation workflows with large preset libraries.
A LoRA and research project built around optical imperfection, anamorphic behaviour and cinematic form in generative systems.
The research argues against the automatic equation of image quality with total visibility. Anamorphic distortion, spatial compression, constrained framing, black-and-white luminance regimes and film-oriented colour are treated as structural cinematic choices rather than defects to be cleaned away.
