docs: rewrite README in Chinese
This commit is contained in:
Marigold1122
@@ -1,71 +1,122 @@
|
||||
# ComfyUI Chat Image Bridge
|
||||
|
||||
ComfyUI custom nodes for image generation through OpenAI-compatible chat APIs,
|
||||
with extra support for Gemini/Nano Banana Pro style image parameters.
|
||||
一个用于 ComfyUI 的第三方 API 生图节点，面向 OpenAI-compatible 聊天接口，同时对 Gemini / Nano Banana Pro 这类图片模型做了额外适配。
|
||||
|
||||
The node does not ship with a built-in provider URL or API key. Users supply
|
||||
their own OpenAI-compatible base URL and credentials at runtime. The default
|
||||
model name is `gemini-3-pro-image-preview` as a convenience placeholder; change
|
||||
it to any model supported by your provider.
|
||||
这个节点不内置任何服务商地址或 API Key。用户需要在节点里填写自己的 `base_url`、`api_key` 和模型名。为了方便上手，默认模型名预填为：
|
||||
|
||||
## Features
|
||||
```text
|
||||
gemini-3-pro-image-preview
|
||||
```
|
||||
|
||||
- OpenAI-compatible `/v1/chat/completions` image generation
|
||||
- `base_url`, API key, and model are user supplied
|
||||
- `Fetch Models` button for OpenAI-compatible `/v1/models`
|
||||
- Optional reference image inputs
|
||||
- `1K`, `2K`, and `4K` resolution controls
|
||||
- Common aspect ratio controls
|
||||
- Gemini native `imageConfig` support for models such as `gemini-3-pro-image-preview`
|
||||
- Decodes Markdown image data URLs, plain image URLs, `b64_json`, and Gemini `inlineData`
|
||||
## 功能
|
||||
|
||||
## Installation
|
||||
- 支持 OpenAI-compatible `/v1/chat/completions` 生图接口
|
||||
- 支持自定义 `base_url`、API Key 和模型名
|
||||
- 支持点击 `Fetch Models` 从 `/v1/models` 获取模型列表
|
||||
- 支持 1K、2K、4K 分辨率选择
|
||||
- 支持常见长宽比选择
|
||||
- 支持 1-2 张参考图输入
|
||||
- 支持解析 Markdown 图片、图片 URL、`b64_json`、Gemini `inlineData`
|
||||
- 对 `gemini-3-pro-image-preview` 等 Gemini 图片模型支持原生 `imageConfig`
|
||||
- 提供高级节点，可输出原始响应和图片引用，方便调试
|
||||
|
||||
Clone this repository into your ComfyUI `custom_nodes` directory:
|
||||
## 安装
|
||||
|
||||
### 启动器安装
|
||||
|
||||
如果你的 ComfyUI 启动器支持从 GitHub 仓库安装自定义节点，可以直接填入本仓库地址：
|
||||
|
||||
```text
|
||||
https://github.com/Marigold1122/Chat-Image-Bridge-Node
|
||||
```
|
||||
|
||||
安装后重启 ComfyUI。
|
||||
|
||||
### 手动安装
|
||||
|
||||
进入 ComfyUI 的 `custom_nodes` 目录：
|
||||
|
||||
```bash
|
||||
cd ComfyUI/custom_nodes
|
||||
git clone <this-repository-url> ComfyUI-ChatImageBridge
|
||||
git clone https://github.com/Marigold1122/Chat-Image-Bridge-Node.git ComfyUI-ChatImageBridge
|
||||
```
|
||||
|
||||
Install optional Python requirements if your environment does not already have
|
||||
them:
|
||||
如果你的环境缺少依赖，可以安装：
|
||||
|
||||
```bash
|
||||
pip install -r ComfyUI-ChatImageBridge/requirements.txt
|
||||
```
|
||||
|
||||
Restart ComfyUI.
|
||||
然后重启 ComfyUI。
|
||||
|
||||
## Simple Node
|
||||
## 基础节点
|
||||
|
||||
Add:
|
||||
添加节点：
|
||||
|
||||
```text
|
||||
api -> Chat Image Bridge -> Chat Image Bridge
|
||||
```
|
||||
|
||||
Fill:
|
||||
主要参数：
|
||||
|
||||
- `api_key`: your provider API key
|
||||
- `base_url`: your provider base URL, such as `https://example.com`, `https://example.com/v1`, or a full `/v1/chat/completions` URL
|
||||
- `model`: the image-capable chat model name
|
||||
- `prompt`: your image prompt
|
||||
- `resolution`: `auto`, `1K`, `2K`, or `4K`
|
||||
- `aspect_ratio`: `auto`, `1:1`, `16:9`, `9:16`, and common ratios
|
||||
- `api_key`：你的第三方中转站或模型服务商 API Key
|
||||
- `base_url`：接口地址，可以填服务商根地址、`/v1` 地址，或完整 `/v1/chat/completions` 地址
|
||||
- `model`：模型名，默认是 `gemini-3-pro-image-preview`
|
||||
- `prompt`：提示词
|
||||
- `resolution`：分辨率，支持 `auto`、`1K`、`2K`、`4K`
|
||||
- `aspect_ratio`：长宽比，支持 `auto`、`1:1`、`16:9`、`9:16` 等
|
||||
- `timeout_seconds`：请求超时时间
|
||||
|
||||
Click `Fetch Models` to query the provider's OpenAI-compatible `/v1/models`
|
||||
endpoint. The node updates the existing `model` dropdown in place.
|
||||
可选输入：
|
||||
|
||||
Connect `image` to `Save Image` or any downstream image node.
|
||||
- `image_1`
|
||||
- `image_2`
|
||||
|
||||
Optional `image_1` and `image_2` inputs send reference images as OpenAI-style
|
||||
`image_url` data URLs.
|
||||
输出：
|
||||
|
||||
## Gemini Image Parameters
|
||||
- `image`
|
||||
|
||||
For Gemini image models such as `gemini-3-pro-image-preview`, non-auto
|
||||
resolution/aspect settings are sent through the native Gemini request shape:
|
||||
可以直接接到 `Save Image` 或其他 ComfyUI 图像节点。
|
||||
|
||||
## base_url 填法
|
||||
|
||||
下面几种都可以：
|
||||
|
||||
```text
|
||||
https://example.com
|
||||
https://example.com/v1
|
||||
https://example.com/v1/chat/completions
|
||||
```
|
||||
|
||||
节点会自动整理为请求需要的接口地址。
|
||||
|
||||
## 获取模型列表
|
||||
|
||||
填好 `api_key` 和 `base_url` 后，点击节点里的：
|
||||
|
||||
```text
|
||||
Fetch Models
|
||||
```
|
||||
|
||||
节点会请求：
|
||||
|
||||
```text
|
||||
/v1/models
|
||||
```
|
||||
|
||||
然后把返回的模型列表填入 `model` 下拉框。
|
||||
|
||||
如果你的服务商不支持 `/v1/models`，也可以直接手动填写模型名。
|
||||
|
||||
## 4K 和长宽比
|
||||
|
||||
对于 Gemini 图片模型，例如：
|
||||
|
||||
```text
|
||||
gemini-3-pro-image-preview
|
||||
```
|
||||
|
||||
当 `resolution` 或 `aspect_ratio` 不是 `auto` 时，节点会优先使用 Gemini 原生请求格式，并传入：
|
||||
|
||||
```json
|
||||
{
|
||||
@@ -79,23 +130,30 @@ resolution/aspect settings are sent through the native Gemini request shape:
|
||||
}
|
||||
```
|
||||
|
||||
If the provider does not support the native Gemini endpoint, the node falls back
|
||||
to OpenAI-compatible `/v1/chat/completions` and keeps the settings as prompt
|
||||
hints.
|
||||
这样 `4K` 和长宽比会作为真正的图片参数发送，而不是只写进 prompt。
|
||||
|
||||
## Advanced Node
|
||||
如果服务商不支持 Gemini 原生接口，节点会自动回退到 OpenAI-compatible `/v1/chat/completions`，并把分辨率和比例作为提示词前缀兜底。
|
||||
|
||||
Use `Chat Image Bridge Advanced` if you need:
|
||||
## 高级节点
|
||||
|
||||
- raw response output
|
||||
- extracted image reference output
|
||||
添加节点：
|
||||
|
||||
```text
|
||||
api -> Chat Image Bridge -> Chat Image Bridge Advanced
|
||||
```
|
||||
|
||||
高级节点适合调试或特殊接口，额外提供：
|
||||
|
||||
- 原始响应 `response`
|
||||
- 图片引用 `image_refs`
|
||||
- `system_prompt`
|
||||
- `size`
|
||||
- custom request fields through `extra_body_json`
|
||||
- up to 14 reference images
|
||||
- `extra_body_json`
|
||||
- 最多 14 张参考图输入
|
||||
|
||||
## Notes
|
||||
## 注意事项
|
||||
|
||||
- `base_url` may be a provider root URL, a `/v1` URL, or a full `/v1/chat/completions` URL.
|
||||
- API keys are entered in the node UI and are not hard-coded in this repository.
|
||||
- Some providers may not support Gemini native endpoints. The node will fall back to OpenAI-compatible chat completions.
|
||||
- API Key 只在节点 UI 里填写，本仓库不会硬编码任何 Key。
|
||||
- 不同中转站对 OpenAI-compatible 的兼容程度不同，字段支持可能不完全一致。
|
||||
- 如果 4K 或比例没有生效，请优先确认服务商是否支持 Gemini 原生 `imageConfig`。
|
||||
- 旧版工作流如果使用过 `endpoint_url`，建议改成新版的 `base_url`。
|
||||
|
||||
Reference in New Issue
Block a user