Chatterbox

LAPLACE Chatterbox's Auto Send loops through preconfigured danmaku templates at a set interval, for live rooms that need the same content repeated over long stretches

Features

• Configure multiple danmaku templates; once started, they are sent in order on repeat
• Each template holds multiple lines, and text past the length limit is split automatically, with every segment sent as its own danmaku
• Random Color: switches at random to one of the danmaku colors unlocked in the current room before each send
• Random Interval: shortens the configured send interval by up to 500ms at random so the pacing looks more natural
• Cloud and local dictionary replacement rules are applied automatically on send
• Templates can contain emote IDs, and each emote is sent as its own message
• Optionally enable YOLO mode: an LLM polishes every non-emote line of the template once before the run starts, and the loop then sends the polished text, with one LLM call per round
• In YOLO mode, switch the active Auto Send prompt directly from a prompt picker inside the panel, without returning to the settings page

Usage

Configure Templates

• Open a bilibili live room and click the "Chatterbox" floating button in the lower right to expand the panel
• Switch to the "Auto Send" tab and fill the template list with the danmaku you want to loop
• Each line counts as one danmaku; text past the max length is split automatically and sent in order

Starting and Stopping

• Click "Start" to begin the automatic loop
• Click "Stop" to halt sending
• While Auto Send runs, the "Chatterbox" floating button turns green and a green 🟢 marker appears on the "Send" tab and in the "Auto Send" panel title
• With YOLO mode on, the "Auto Send" panel title also shows a ⚡️ marker

YOLO Mode

• Toggle it with the "YOLO" button below the template editor: filled means enabled, outlined means disabled
• With YOLO on, one LLM call polishes every non-emote line before each round, and the polished text then goes through the normal send flow; emote lines are never rewritten
• Set API Base, API Key, and the model under "Settings → LLM Settings", then activate a prompt under "Settings → LLM Prompts → Auto Send"; until both are done, the YOLO button stays disabled
• When YOLO is on and the LLM is configured, a prompt dropdown appears to the right of the YOLO button, so you can switch the active Auto Send prompt without leaving the panel
• If the LLM call fails, that round falls back to the raw template text and keeps running

Options

Notes

• Random Color depends on the room's danmaku color configuration, and only colors the current account has unlocked enter the random pool
• Auto Send goes through the shared send queue and honors bilibili's limit of one message per second; when a manual send (regular send, one-click meme send) fires, queued Auto Send messages give way to it
• The on/off state can be remembered per live room: check "Keep Auto Send State for This Live Room" and it resumes automatically the next time you enter that room
• YOLO polishing finishes before each round begins and counts as whole-round overhead rather than time inserted between individual danmaku, so pacing matches the original template; a multi-line template makes one LLM call per line
• YOLO only touches plain-text lines; emote ID lines in the template are always sent as-is and never rewritten by the LLM