Photoshoot for ComfyUI: A Series of Frames of One Character Without LoRA

A set of nodes ralksta/ComfyUI-Photoshoot (MIT license) is open on GitHub, which, based on a single 44-field character description, assembles a deterministic series of up to 40 frames with different angles, poses, and emotions — using pure prompt engineering, without LoRA and reference photos.

Compiled by Sergey KostenchukPublished 2026-09-15Updated 2026-09-18

2026-09-15 Multimodal

🎨 Photoshoot for ComfyUI: A Series of Frames of One Character Without LoRA

In the ralksta/ComfyUI-Photoshoot repository (MIT, ~78 stars), the character is described once in 44 fields and saved under a name; the node assembles up to 40 frames along six axes: camera, pose, expression, focus, format, noise. The series is deterministic and can be extended later without repeats.

🌍 Stability — through pure prompt engineering: in a wide shot, the description is compressed, because details unresolvable from a distance pull the frame to the face. The nodes write only text, so they are compatible with Flux, SD 3.5, Qwen-Image, Krea 2; with SD 1.5 and SDXL, they are weaker. There are no identity measurements in the repository.

👤 Installed via ComfyUI Manager (search for "Photoshoot"). Inside — the workflow photoshoot-series.json, examples, and a guide; in version 2.3.0, Styles and Monochrome nodes were added. Posters in the documentation were rendered on Krea 2 Turbo (8 steps, CFG 1.0).

Source 1: https://github.com/ralksta/ComfyUI-Photoshoot

Sources