CoreWise.video

Deep Research -- Full long-form research articles from a single question -- 4-layer Mixture-of-Agents pipeline: parallel web search (deep search + YouTube + Perplexity), 3 reasoning models, 2 revision models, final synthesis -- Inline citations linking to every source used -- Search filters: language, date range, academic-only sources -- Prompt optimizer to sharpen your question before the pipeline runs -- Organize articles into projects Mind Maps -- Turn any deep research article into an interactive visual mind map -- Sections branch into subtopics with spatial connections Research Slides -- Generate a full presentation from any deep research article -- 12 style presets across dark and light themes -- Title slides, content slides, quote callouts, two-column comparisons, FAQ slides, sources slide -- Downloads as a single HTML file — present from anywhere with no dependencies -- Quick-access Slides button in the library listing and export dropdown Voice Input -- Microphone button on every major text input (deep research, extraction Q&A, synthesis Q&A) -- Chrome/Edge/Safari use native Web Speech API -- Firefox gets Vosk — an offline speech model running entirely in-browser via WebAssembly (no audio leaves your machine) Model Council -- Ask the same question to 2–6 AI models simultaneously from 16+ providers (Claude, GPT, Gemini, DeepSeek, Llama, Mistral, Qwen, and more) -- Responses stream in side by side so you can spot agreements, contradictions, and unique angles -- GPT 5.2 runs a comparative synthesis: convergence table, divergence breakdown, unique discoveries, and a bottom line -- Follow-up questions build on the full multi-model context -- Web search toggle gives models live internet context -- Save your preferred model combination as a default Debate Arena -- Models don't just answer — they evaluate each other's responses with honest critiques -- Multi-round debates with thinking indicators as models deliberate -- Arbiter (GPT 5.2) delivers a structured verdict: consensus findings, research scorecard, blind spots, unresolved questions -- Debate insights route into the Synthesis Engine for further analysis -- Debate content persists to database — come back to it anytime Text Highlighting & Notes -- Select any text in an extraction or research article and pick from six highlight colors -- Highlights persist permanently across sessions -- Notes drawer with Markdown support and auto-save -- Highlight a passage and send it directly to your notes Creative Domain Focus -- New extraction mode built for creative content (filmmaking, music, writing, design) -- Captures creative philosophy, techniques, process, key tensions, and inspiration — not generic bullet points -- Set per-extraction or as your default in Profile Settings -- Four domain modes total: All Purpose, Research Paper, Tutorial, and Creative X (Twitter) Support -- Paste any tweet URL for full multi-model extraction -- Video tweets are transcribed, text tweets are analyzed -- Same extraction pipeline as YouTube, articles, and PDFs Mistral OCR for Scanned PDFs -- Automatic fallback for PDFs with no text layer (scanned docs, photographed handouts) -- Detects when standard extraction returns near-empty results and switches to Mistral vision OCR Achievements System -- Added 6 starter achievements to guide feature discovery -- Used to gain quicker access to CORE+ features -- Perplexity Sonar Deep Research - After extracting a video/article/PDF, run Sonar Deep Research to get a sourced analysis with cited sources. The Q&A then draws from both the original extraction AND the deep research. -- Grok & Mistral as extraction providers -- Extract Later - Bookmark/save content on the Discover page for later extraction. Persistent playlist at the top of Discover. -- Web Search on Discover - "Web" tab powered by Perplexity with AI-synthesized results + source citations. Discover 5 sources: YouTube, TikTok, Instagram, Reddit, Web. -- Research Paper domain focus — Specific system prompt and formatting for this domain focus; best with research reports -- "Synthesized by" badge — Every extraction now shows which synthesis engine produced the result. -- Spotify Podcast Support | Auto-finds RSS feed, matches episode, pulls direct audio for transcription. Reliability -- Fixed occasional OOM crashes -- PDF service startup timeout extended with early crash detection -- Captionless video detection and fallback system

Browser Extensions -- Firefox addon now available (in addition to Chrome extension) -- Extract wisdom on YouTube with one click Multi-Platform Discovery -- Search YouTube, TikTok, Instagram, and Reddit directly from CoreWise to extract wisdom Multilingual Support -- Generate audio summaries in 20 different languages -- Entire platform now available in your preferred language -- Convert to podcast format in your preferred language More AI Models (available to all users) -- Llama 4 Scout Instruct and Llama 4 Maverick Instruct now available to all users -- Gemini 3 Pro Preview available -- Choose between different synthesis engines for your preferred AI style of the final step AI Quality Improvements -- Significantly optimized synthesis prompts across all AI models for better output quality and readability -- Improved long-form video transcript handling with automatic fallback systems -- Fixed bug with incorrect model attribution in Unique Model Perspectives Smarter PDF Handling -- Extract wisdom from academic papers (including arXiv) without needing to end in .pdf -- More reliable processing for all PDF types Full Guide Mode -- New extraction format perfect for tutorials and how-to videos -- Get complete step-by-step breakdowns -- Optimized for educational content Extraction Speed -- Significantly faster processing times -- Smarter caching means instant results for popular videos Reliability Improvements -- Replaced direct media transcription engine for higher quality audio/video processing -- Added automatic retry logic to prevent transient network failures -- Smart rate limit handling with CORE+ priority processing -- Improved database stability and connection management Share Your Wisdom -- Share any extraction with a simple link without needing to share to Community page (bug fixed) -- No signup required for viewers -- Works automatically with social media sharing Better Guidance -- Added validation warnings for YouTube channels and platform homepages to prevent errors -- Improved extraction button layout and animations for better visibility -- Updated extraction loading sequence text for better clarity Mobile Experience -- Completely redesigned for phones and tablets -- Easier navigation and better button placement -- Full-featured mobile discovery Visual Polish -- Improved dark mode and night-light themes -- Better readability everywhere -- Smoother animations and transitions Faster Loading -- Optimized initial page load and snappier performance overall -- Code splitting and font optimization CORE+ Updates -- Real-time support chat - help when you need it directly from the developer (not AI) Fixes -- Markdown formatting improvements -- Notion exports handle long text without issue -- Dozens of small improvements to reliability

Initial release of CoreWise.video.