AskAvatar

Initial Release
AskAvatar is an AI-powered streaming companion designed to bring customizable mascots to life. It generates context-aware, reactive responses to stream events, encouraging chat interaction and giving streamers a unique, interactive prop.

Core Features

• 13 Pre-Built Avatar Profiles: Ready to use out of the box.
• Reactive Animation System: Avatars utilize a 4-layer PNG system (background, closed mouth, open mouth, wide-open mouth). The software dynamically switches between these layers and applies bounce animations synced to the generated audio waveform to simulate speech.
• Full Customizability: Users can swap all four image layers and completely customize the avatar's core personality instructions.
• AI Voices: Includes 13 royalty-free English voices powered by Piper AI, with instructions for installing additional languages.
• Animations: Various enter and exit animations for the avatar.

Integrations & Modes

• Stream API Support: Twitch and YouTube events are relayed to the avatar via Streamlabs and StreamElements API socket keys.
• VTuber Mode: A toggleable override that pauses AI event reactions and allows the avatar to function as a standard reactive PNG VTuber driven by microphone volume.

Tech Stack
Built utilizing ChatGPT, Gemini, Ollama (AI Responses), and Piper (AI Voices).

Platform: Uploaded to the Microsoft Store with a trial version enabled.

Bug Fixes:

• Fixed a packaging issue that prevented the app from locating the avatar image asset library.
• Fixed an issue where avatars lingered on screen after finishing a response instead of progressing to the next event in the queue.

• Added merchandise purchases to the alert options.
• Implemented the first iteration of outgoing webhook integration.
• Minor UI cleanups and visual rearrangements for the avatar images.

• Added 4 new AI voices.
• Enabled repositioning and rescaling of the overlay window.
• Added Chroma Key support for enhanced compatibility with OBS and other broadcasting software that may struggle with transparent windows.

Bug Fixes:

• Piper AI audio files now auto-delete upon app closure, rather than storing for 30 days.
• Improved overlay window stability to prevent it from getting stuck in an "off" state.

• Event Triggers (v1): Introduced conditional image swapping. Users can now assign specific avatar images to distinct alerts or tip amounts (e.g., a unique reaction image specifically for subscriptions).
• Added an auto-check and user notification system to ensure Ollama (a required third-party app) is installed.
• Added a toggleable guide border to the overlay window to assist users in locating and positioning it.
• AskAvatar Adult Edition Released: A standalone variant tailored for Chaturbate. Features 9 unique NSFW avatars, restricted tipping context (tip messages are not forwarded to the AI to protect privacy), and adult-themed personalities.

• Event Trigger Unique Scenarios: Users can now inject temporary prompt modifiers for specific events. (e.g., instructing the avatar to act like a mosh pit started during a raid, or reacting specifically to a $4.20 tip).
• Added Pitch and Rate controls for AI voices.
• Added bounce animations and enhanced mouth reactivity.
• Implemented the first iteration of profile exporting and importing.
• Added a "Check for Updates" button to the Integration page.

Bug Fixes:

• Corrected currency type formatting in AI responses.
• Improved Ollama detection and added background auto-run instructions.

Tools Used: Copilot, Antigravity

• Added foreground layering support.
• Added AI model selection.
• Backend Overhaul: Decoupled image files from the master profile JSON. This improves backup stability, ensures persistent data during updates, and allows profiles to be exported/shared as a single file without hard-linking local images.

AskAvatar Adult Updates:

• Imported the amended Event Triggers, profile import/export, and foreground layers from the main branch.
• Added instructions for downloading less restrictive AI models via Ollama to better fit the app's context.

Major Additions

• Near 200 new media assets added including a new avatar (Angelika).
• TikTok Live Integration (v1): Full support added, including custom TikTok thresholds.
• Kick Integration (v1): Full support added, including custom Kick thresholds.
• The Stage: A complete overhaul of the overlay system. Users can now position multiple event triggers and the default avatar at various scales and locations across a dedicated, auto-scaling canvas. Features layer highlighting for easy management and multi-monitor support. (Set to auto-open by default).
• Video Layers: Added foreground and background video support with controls for volume, delay, and continuous looping.
• Vibe Checker: Links Twitch/TikTok chat to the AI, feeding it the last 20 messages for context when formulating responses. Includes a username blocklist to filter out bot announcements or specific users.

New Modes & Overrides

• No Event Info Mode: Skips the username and event type announcement, jumping straight to the AI response.
• No Riff Mode: Skips the AI response generation, allowing the avatar to simply read the event info.
• Persona Override (Sub-Avatars): Temporarily swaps the avatar's personality during specific Event Triggers, allowing for multiple character personas within a single profile.
• Video Only Mode: Triggers a video overlay without generating an AI response.
• Simultaneous & Randomized Videos: Event Triggers can now select a random video from a designated pool, or play multiple triggered videos simultaneously (e.g., dropping dozens of feather animations during a mass raid).
• Advanced Overrides: Customize voice, pitch, rate, reactivity, bounce, response length, temperament, and enter/exit animations on a per-trigger basis.

Voice & LLM Expansions

• In-App Voice Management: Browse, download, and auto-install 15+ English AI voices directly within AskAvatar, alongside easy access to 45+ HuggingFace language packs.
• Third-Party API Support (Beta): First iteration of support for ElevenLabs and OpenAI TTS (requires user subscription).
• Third-Party LLM Support (Beta): Option to route AI responses through OpenAI (ChatGPT), Anthropic (Claude), or Gemini instead of Ollama.
• Updated default AI model with backend instructions for optimized Ollama downloading.

System Controls & UI

• Complete UI overhaul with customizable themed colors.
• Added a Help & Settings side panel with tutorials and vital performance controls.
• CPU Thread Limiter: Control how many CPU cores the AI utilizes for generation.
• Context Limiter: Manage RAM/VRAM allocation to keep Ollama and Piper running in the background, preventing cold-boot lag for every response.
• Fast Sentence Streaming: Enabled by default. Passes text to the voice generator as it is actively being written by the AI, significantly increasing performance speed.
• Added information hover cards across the UI for instant tooltips.
• Added Avatar temperament and response length sliders.
• Added Profile renaming and a "Revert to Default" template loader for personalities.
• Export Trigger Menu: Generate a single image detailing all active thresholds and event triggers to easily share with viewers on stream profiles.
• Note: Exported profiles currently only store file directory paths for videos. Users must back up video assets separately and re-link them when migrating profiles.

Major Additions

• 1-Click Installation: Ollama AI is now packaged within the initial installer, making AskAvatar a seamless 1-click install program.
• AI Model Manager: Added an Ollama AI model selection and download window, allowing users to select from a list of AI models to use for their avatar personality, with an option to add custom AI models.
• Kokoro TTS Engine: A second, higher-quality Text-To-Speech engine has been added in addition to Piper. Kokoro adds 28 additional English AI voices recorded at a studio level, as well as additional international voices.
• Browser Source Support: Users can now have The Stage overlay displayed within OBS via a Browser Source instead of a second window, increasing performance.
• UI Localization: Added UI language change options for Spanish, German, Japanese, Russian, Portuguese, French, Korean, and Chinese.

Bug Fixes

• Profile Refresh fixed; Import and Export functions corrected to include additions to event triggers and the stage overlays.
• Resizing of event trigger transparent video assets to 1080p and 720p to increase performance at no loss to image quality.