Fake Twitch Chat

Fake Twitch Chat injects a live, AI-powered Twitch-style chat sidebar into any webpage. The chat reads the content of whatever you are looking at and generates realistic, chaotic, and occasionally hilarious fake Twitch chat messages reacting to it in real time. FEATURES AI-Powered Chat Generation The extension sends page context (title, description, and site-specific content) to the selected AI model, which generates batches of authentic-feeling Twitch chat messages. Messages follow real Twitch chat patterns: short bursts, emote spam, bandwagon reactions, recurring personalities, and the occasional unhinged off-topic comment. Six Chat Personality Modes Switch between fundamentally different chat personalities from the popup: Default - Mixed crowd: hype, toxic, wholesome, and bad takes all at once Hype Train - Pure unfiltered excitement. Everything is incredible. W W W. Toxic - Sarcastic, mocking, never impressed. Skill issue. Ratio. Wholesome - Supportive and kind. You got this. W human being. Backseaters - Everyone knows better. Unsolicited advice, maximum confidence. Clueless - Every chatter is confidently wrong about something completely different. No shared reality. Intensity Slider A 0-10 slider controls how strongly the selected personality comes through. At 0 it is almost indistinguishable from normal chat. At 10 it is fully unhinged with nothing held back, but always reacting to what is actually on the page. Three Tiers Free - Hardcoded message pools per mode. No API key needed. No AI. Standard - AI-powered. Fetches 30 messages at a time, refetches every 40s. Approximately $0.15-0.18/hr with Anthropic. Turbo - Smaller batches fetched every 10-12s. Chatters call each other out by name, argue, agree, and build running threads. Approximately 3x the cost of Standard, driven by fetch frequency. Multi-Provider AI Support Choose between Anthropic (Claude Haiku) and OpenAI (GPT-4o mini) from the Settings page. Both keys are saved separately so you can switch providers without re-entering credentials. OpenAI is approximately 5x cheaper than Anthropic for the same usage. BTTV Emote Rendering Emote names in chat messages are replaced with real animated images sourced from the BetterTTV CDN. Includes KEKW, OMEGALUL, Pog, Sadge, monkaS, and many more. Global emotes are fetched once and cached for 24 hours. Chat Follows You Across Navigation Enable the chat once and it stays with you as you browse. On single-page apps (YouTube, Reddit, Twitter/X) a subtle separator line appears in the chat when you navigate. On full page navigations the chat auto-enables on the new page using your saved settings. You Can Talk to the Chat A message input bar sits at the bottom of the sidebar. Type something and press Enter. Your message appears immediately with a purple accent and your streamer name, and the next AI batch reacts specifically to what you said. Smart Page Context The extension has site-specific extractors that pull the most relevant content for each context. On video pages it reads the title, channel, and current chapter. On social media it reads post text and top comments. On code repositories it reads the project name and README. On streaming pages it reads the stream title and game. On encyclopedia articles it reads the title and opening paragraph. On everything else it reads the page title and article body, falling back to the meta description. Per-Site Settings Mode, tier, intensity, and font size are saved per hostname. Visiting YouTube will restore your YouTube settings; Reddit restores your Reddit settings. Sidebar Controls - Collapse the sidebar to a 36px strip. Chat pauses while collapsed so no tokens are spent in the background. - Drag to resize between 220-520px wide. Width is saved to storage. - Pause and resume the message drip without disabling the extension. - Dark and light theme toggle, synced between the sidebar and popup. - Clip button takes the last 12 messages and renders them to a shareable PNG card. - Keyboard shortcut Alt+T cycles the sidebar: off, active, collapsed, active. Tab Visibility Awareness The extension pauses fetching and the message drip when the tab is in the background, and resumes automatically when you switch back. Usage Tracking The popup shows estimated cost, total API calls, and total tokens used since the last reset. SECURITY AND PRIVACY This extension reads visible page content and sends it to an external AI API. That is its entire job, and it is designed to be fully transparent about it. Four-layer protection model: 1. Sensitive site blocking - A hardcoded list of known sensitive hostnames is blocked by default. This includes major email providers, payment services, password managers, and government portals. Any domain using a .gov, .bank, or .mil top-level domain is also blocked universally. 2. Custom blocked sites - Add your own list of hostnames in Settings that should always show the sensitive site warning. 3. Whitelist mode - Locks the extension to a user-defined list of approved hostnames only. The enable toggle is fully disabled on every other site. 4. Per-site consent gate - Before the very first API call on any hostname, the sidebar shows a preview of exactly what will be extracted and which AI provider it goes to. You approve once per site and are never asked again. API keys are stored in chrome.storage.local, sandboxed to the extension, and only ever sent to the selected provider's API endpoint. The free tier makes no API calls and no data ever leaves the browser. COST REFERENCE Anthropic (Claude Haiku): - Standard, 1 hour: approximately $0.15-0.18 - Turbo, 1 hour: approximately $0.45-0.55 OpenAI (GPT-4o mini): - Standard, 1 hour: approximately $0.03-0.04 - Turbo, 1 hour: approximately $0.08-0.12 Actual costs vary based on page content length and history size. The usage tracker in the popup shows your running total.