# Krea2 — 8-Step Cinematic Portrait Baseline (Episode 272)
ComfyUI workflow exported from **AI-KSK episode 272**.
**File:** `Krea2｜8步电影级人像基准.json`
## What it does
Krea2 — 8-Step Cinematic Portrait Baseline. The episode walks through the graph node by node,
explains the model choices, and shows sample outputs.
## How to run
1. Open ComfyUI.
2. Drag `Krea2｜8步电影级人像基准.json` onto the canvas (or *Workflow → Open*).
3. Install any missing nodes the ComfyUI Manager reports.
4. Put the models the graph references into your `models/` folders.
5. Queue the prompt.
Run it without a local GPU on RunningHub:
https://www.runninghub.cn/post/2072175228620464130/?inviteCode=rh-v1111
## Notes
- Tutorial workflow, not a packaged product. Parameters are those used in the episode.
- Model weights are not bundled; each model keeps its own licence.
- Walkthrough video (Chinese): https://youtu.be/zByZX0dN9Ns
- Bilibili: https://www.bilibili.com/video/BV1FWT76KENn/
- Independent AI-KSK release. Not affiliated with or endorsed by the model authors, Comfy Org
or RunningHub.
ComfyUI workflow exported from **AI-KSK episode 272**.
**File:** `Krea2｜8步电影级人像基准.json`
## What it does
Krea2 — 8-Step Cinematic Portrait Baseline. The episode walks through the graph node by node,
explains the model choices, and shows sample outputs.
## How to run
1. Open ComfyUI.
2. Drag `Krea2｜8步电影级人像基准.json` onto the canvas (or *Workflow → Open*).
3. Install any missing nodes the ComfyUI Manager reports.
4. Put the models the graph references into your `models/` folders.
5. Queue the prompt.
Run it without a local GPU on RunningHub:
https://www.runninghub.cn/post/2072175228620464130/?inviteCode=rh-v1111
## Notes
- Tutorial workflow, not a packaged product. Parameters are those used in the episode.
- Model weights are not bundled; each model keeps its own licence.
- Walkthrough video (Chinese): https://youtu.be/zByZX0dN9Ns
- Bilibili: https://www.bilibili.com/video/BV1FWT76KENn/
- Independent AI-KSK release. Not affiliated with or endorsed by the model authors, Comfy Org
or RunningHub.
Description
# Krea2 — 8-Step Cinematic Portrait Baseline (Episode 272)
ComfyUI workflow exported from **AI-KSK episode 272**.
**File:** `Krea2｜8步电影级人像基准.json`
## What it does
Krea2 — 8-Step Cinematic Portrait Baseline. The episode walks through the graph node by node,
explains the model choices, and shows sample outputs.
## How to run
1. Open ComfyUI.
2. Drag `Krea2｜8步电影级人像基准.json` onto the canvas (or *Workflow → Open*).
3. Install any missing nodes the ComfyUI Manager reports.
4. Put the models the graph references into your `models/` folders.
5. Queue the prompt.
Run it without a local GPU on RunningHub:
https://www.runninghub.cn/post/2072175228620464130/?inviteCode=rh-v1111
## Notes
- Tutorial workflow, not a packaged product. Parameters are those used in the episode.
- Model weights are not bundled; each model keeps its own licence.
- Walkthrough video (Chinese): https://youtu.be/zByZX0dN9Ns
- Bilibili: https://www.bilibili.com/video/BV1FWT76KENn/
- Independent AI-KSK release. Not affiliated with or endorsed by the model authors, Comfy Org
or RunningHub.
ComfyUI workflow exported from **AI-KSK episode 272**.
**File:** `Krea2｜8步电影级人像基准.json`
## What it does
Krea2 — 8-Step Cinematic Portrait Baseline. The episode walks through the graph node by node,
explains the model choices, and shows sample outputs.
## How to run
1. Open ComfyUI.
2. Drag `Krea2｜8步电影级人像基准.json` onto the canvas (or *Workflow → Open*).
3. Install any missing nodes the ComfyUI Manager reports.
4. Put the models the graph references into your `models/` folders.
5. Queue the prompt.
Run it without a local GPU on RunningHub:
https://www.runninghub.cn/post/2072175228620464130/?inviteCode=rh-v1111
## Notes
- Tutorial workflow, not a packaged product. Parameters are those used in the episode.
- Model weights are not bundled; each model keeps its own licence.
- Walkthrough video (Chinese): https://youtu.be/zByZX0dN9Ns
- Bilibili: https://www.bilibili.com/video/BV1FWT76KENn/
- Independent AI-KSK release. Not affiliated with or endorsed by the model authors, Comfy Org
or RunningHub.