Seedream 5.0 Pro: Image Edit is a precision image-to-image and inpainting workflow that lets you change exactly what you want while keeping the original lighting, depth, and texture intact. It starts with LoadImage to bring in your source photo, then optionally uses Painter to create a clean mask for targeted edits. The core transformation happens in ByteDanceSeedreamNodeV2, which conditions on your input image (and mask if provided) to apply prompt-driven changes without destroying structure. Finally, SaveImageAdvanced writes out clean results with controllable filenames and folders.
Technically, the mask from Painter constrains edits to specific regions, while the input image conditioning helps Seedream preserve scene geometry and materials. You’ll get two outputs: a main edited image from ByteDanceSeedreamNodeV2 and a second preview that reflects the Painter-guided variant—useful for quick QA and before/after comparisons. Tuning typical parameters like prompt/negative prompt, denoise strength (how much to change vs. preserve), steps, guidance (CFG), and seed lets you dial in subtle retouches for product shots, keep consistent faces across a series, or lay out multilingual text in a structured composition.
API
코드에서 이 워크플로우 사용하기
모든 Comfy 워크플로우는 JSON 그래프입니다. 아래 페이로드는 이 워크플로우 그 자체로, ComfyUI에서 실행되는 것과 동일합니다 — URL에서 가져오거나, 버전 관리에 보관하거나, ComfyUI에서 불러와 노드별로 실행할 수 있습니다.
Comfy SDK로 TypeScript 또는 Python에서 실행하세요. 동일한 코드는 Comfy Cloud 또는 직접 호스팅하는 ComfyUI 모두에 적용됩니다 — 기본 URL만 다릅니다.
// Install (beta)
npm i @comfyorg/sdk
// Run this workflow (TypeScript)
import { Comfy } from "@comfyorg/sdk";
const client = new Comfy({ apiKey: "comfyui-..." });
const wf = await client.workflows.fromFile("workflow_api.json");
const job = await client.run(wf);
await job.getOutputs("<output-node-id>")[0].toFile("output.png");SDK는 API 형식의 워크플로우를 사용합니다: 이 워크플로우를 ComfyUI에서 열고 파일 → 워크플로우 내보내기(API)를 사용하세요.
FAQ
