Create hyperrealistic, identity-consistent AI characters from one reference image.

Who it's for: creators who want this pipeline in ComfyUI without assembling nodes from scratch. Not for: one-click results with zero tuning — you still choose inputs, prompts, and settings.

Open preloaded workflow on RunComfy

Open preloaded workflow on RunComfy (browser)

Why RunComfy first
- Fewer missing-node surprises — run the graph in a managed environment before you mirror it locally.
- Quick GPU tryout — useful if your local VRAM or install time is the bottleneck.
- Matches the published JSON — the zip follows the same runnable workflow you can open on RunComfy.

When downloading for local ComfyUI makes sense — you want full control over models on disk, batch scripting, or offline runs.

How to use (local ComfyUI)
1. Load inputs (images/video/audio) in the marked loader nodes.
2. Set prompts, resolution, and seeds; start with a short test run.
3. Export from the Save / Write nodes shown in the graph.

Expectations — First run may pull large weights; cloud runs may require a free RunComfy account.

Overview

Build hyperrealistic, identity-consistent AI characters from a single reference image. Consistent Character Creator 3.8 expands the character pack with turnarounds, scene variants, close-ups, try-on paths, and dataset export utilities for fast, consistency-focused iteration inside one canvas.

Important nodes:

Key nodes in Comfyui Consistent Character Creator 3.8 workflow

• TextEncodeQwenImageEditPlus (#1199). Feeds Qwen with your reference image plus a short instruction so the model knows what to change and what to keep. Keep image1 connected to your character photo and write concise prompt text when you need pose or scene changes rather than style rewrites. The less you over-describe identity, the more consistent the outputs remain.

• FluxKontextMultiReferenceLatentMethod (#1351). Locks identity features early in the sampling schedule so the subject survives across angles and lighting. Use it as provided for strongest identity; relax only if you intentionally want more restyling.

• KSampler (#199). Converts conditioning into images across all packs. If you need more fidelity, adjust steps; if you want looser adherence to the instruction, adjust cfg. Small coordinated changes here affect every group in predictable ways.

• LoraLoaderModelOnly (#547). Attaches the Qwen-Image-Edit-2511-Lightning LoRA to the UNet. If you see oversharpening or the look drifts, slightly lower strength_model; raise it if edits feel too weak. This is the main speed-to-quality dial for Qwen edits. lightx2v/Qwen-Image-Edit-2511-Lightning

• USOStyleReference (subgraph). Combines CLIP‑Vision features with FLUX via the USO projector to preserve your pack’s style while upscaling for dataset export. If faces crop too tight or too loose, adjust the crop option in the CLIP‑Vision encoder for more stable framing. Comfy-Org/USO_1.0_Repackaged

• Florence2Run (#1413). Auto‑generates captions used by the dataset writer. Leave the caption task on detailed mode for training datasets; switch to shorter modes only if you need compact tags. gokaygokay/Florence-2-Flux-Large

• SaveImageTextDataSetToFolder (#1440). Writes images and paired captions into a clean folder using your character name. Set folder_name to keep datasets separate per character and filename_prefix for reproducible naming.

Notes

Consistent Character Creator 3.8 in ComfyUI | Hyperrealistic Consistent AI Characters — see RunComfy page for the latest node requirements.