Scrapyard AI by Indie

Scrapyard.AI redesigns Character.AI with near-black surfaces, olive accents, Lexend typography, and a compact navigation rail. It styles discovery pages, profiles, character creation, account menus, and chats while preserving Character.AI’s existing controls.
- Site-wide interface: Poster-style discovery shelves, character grids, recent-chat shortcuts, Character Bank, community discovery, and search. Profile sorting and visibility controls flank the category tabs.
- Account menu: The bottom-left Your Profile button opens Character.AI’s native account menu, including profile links, Settings, policies, and logout, styled to match the extension.
- Chat layout: User and character messages align left, with bubbles expanding to fit their text. Message menus stay anchored to each bubble’s top-right corner.
- Growing message composer: A centered input expands as you type, stops growing at a height limit, and shrinks to fit smaller screens.
- Instant text display: Shows text as soon as the client receives it, bypassing the visual typing delay. This changes display speed; it does not accelerate model generation.
- Text styling: Italics receive a blue-gray treatment, while quoted dialogue uses a warm cream color that takes priority over italic coloring.
- Higher-quality portraits: Uses Character.AI’s higher-resolution /i/400 images where available. The sidebar portrait preserves its proportions without added letterboxing.
- Separate portrait sizing: Resize user and bot portraits independently from 24–160 pixels. Portraits expand into available space before shifting the message column.
- Saved image replacements: Upload replacement user and character portraits. Your portrait applies across chats; character replacements save separately for each character and also update its sidebar image. Changes affect your browser only.
- Reading controls: Adjust font size, line spacing, maximum message width, and typeface. Includes higher-contrast and reduced-motion options.
- Reply shortcuts: Use Left and Right arrows to browse generated replies. Right generates another reply when you reach the end. Shortcuts preserve normal text-editing behavior.
- Character sidebar: View the character portrait, publisher, selected model, character profile, and model controls.
- Permission-aware character information: Edit Character opens the native editor when you have permission. Otherwise, it opens a custom information panel containing available descriptions, greetings, definitions, metadata, and character settings. Private definitions are explicitly marked hidden.
- Local Ollama models: Generate replies using installed Ollama models, defaulting to http://localhost:11434. Includes model selection, streaming, stop, retry, regenerate, and copy controls.
- Character-aware local generation: Supplies Ollama with available character information, additional instructions, and conversation context. Context length is adjustable from 2–200 messages; hidden definitions remain unavailable.
- Local conversation storage: Keeps local exchanges and generated alternatives across reloads, with up to 20 alternatives for the latest local exchange.
- Optional Character.AI history saving: Attempts to save completed local exchanges into native Character.AI history. Includes save status, retries, and ordered saving of pending exchanges.
- Generation diagnostics: Displays provider, model, generation state, response timing, and character count. Ollama additionally supplies reported token counts, generation speed, and model-loading time.
- Persistent preferences: Appearance, portrait replacements, reading options, and model settings save locally. The bottom-right gear opens extension settings.
- Startup and performance improvements: Applies the theme early to reduce flashes of the original interface, removes unused sidebar spacing, and avoids repeatedly scanning unchanged pages.
- Privacy: No analytics or credential extraction. Ollama requests go to the selected loopback server; enabling history saving also sends those exchanges to Character.AI.