LmBox Bridge

English

简体中文

English

简体中文

Appearance

AI helper ​

Overview ​

An AI chat tool in LmBox Bridge, built on OpenAI-compatible APIs. It works with OpenAI, Alibaba Cloud Qwen and other compatible providers, and helps with Premiere Pro and After Effects questions, expression writing and more. Members can use the built-in models; everyone else can add their own API key.

Interface ​

AI helper chat

How to open ​

On the extension home page, click the button at the bottom center, then click in the sidebar.

How to use ​

Please note

Running large AI models is an ongoing cost, so long-term availability cannot be guaranteed. LmBox Bridge members can use the built-in models; if you are not a member, set up your own API key to keep using the tool.

Top bar ​

  • Set your local API key and turn the local model on or off. Details
  • Open and manage the list of past chats
  • Switch the OpenAI model
  • Start a new chat

Buttons in code blocks ​

TIP

Code blocks are created on the fly, so copying or running code sometimes fails. Scroll the mouse wheel a little and try again.

  • Run the script in Premiere Pro or After Effects
  • Write the expression to the selected layer property in After Effects
  • Copy the code block

Buttons in the chat ​

  • Edit a question again (clicking it removes all the messages after it)
  • Send what you typed, or press Shift + Enter
  • Stop the reply

Set up your own chat ​

  • The Use the current key switch turns the online key and your local key on or off.

Please note

Non-members can only use a key they set up locally. Members get the models we provide automatically and can start right away.

  • apiKey is the key every large language model needs.
  • baseURL is the server address. Presets are included for OpenAI, Qwen (Alibaba Cloud) and several relay providers.
  • If you have your own OpenAI account, get a key from the OpenAI sign-up page.
  • Alibaba Cloud's Qwen models also work well and come with a generous free quota. Get a DashScope API key (Alibaba Cloud help page).
  • Third-party relay providers that resell OpenAI-compatible access also work: enter the baseURL and key they give you.
  • Context messages lets the model take earlier messages into account. More context gives better answers but uses more tokens; for simple chats, 3 to 5 messages are usually enough.

Video tutorial ​

Video coming soon

Release notes ​

  • 2.7.0 When you create a new Keychain you can pick common providers such as Kimi, SiliconFlow, StepFun, iFlytek Spark, xAI, Mistral and Groq; the address fills in once you choose a platform.
  • 1.6.0 Chat and agent modes can use "Official cloud / Golden key", so you do not need your own model key.
  • 1.6.0 Image and video modes can generate with the official Golden key (your account needs that capability turned on).
  • 0.2.5 You can set a different large model for chat, images, video, audio and other uses, and pick one on this page as needed.
  • 0.2.0 "Chat", "Agent" and "Image generation" each remember the Keychain and model you used last, and switch back when you change mode.
  • 0.1.8 Fixed an issue where the agent sometimes stopped replying, or the screen got stuck, during follow-up questions inside an editing extension that needed the extension's own tools.