Content And Multimedia Tools

This page is a flat, growing list of AI tools for creating and processing media. Every entry is a link followed by a short description of what the tool does and which stack it uses. The tools cover speech synthesis and recognition, voice cloning, OCR for books and handwriting, video and audio downloading, dubbing, transcription, meeting notes, design canvases, and image editing. Some entries point to dedicated notes elsewhere in the vault, such as the voice cloning comparison and NovelKit. A few entries carry caveats, including license limits and a watermark detector that only finds one kind of watermark. Use it as a starting point when you need a tool for one media task, then follow the links for depth.

• https://github.com/izwi-ai/izwi — Local-first audio inference engine for TTS, ASR, speaker diarization, voice cloning, and voice design.
• https://github.com/averygan/reclip — Self-hosted, open-source video/audio downloader with a clean web UI; supports 1000+ sites via yt-dlp.
• https://github.com/duixcom/Duix-Avatar — Open-source AI avatar toolkit for offline video generation and digital human cloning with voice and appearance cloning.
• https://github.com/thainph/node-trans — Real-time audio transcription and translation app (web/Electron); supports cloud Soniox and local Whisper with speaker diarization.
• https://github.com/vorojar/Folio-OCR — Three-pane OCR workbench for books/documents based on GLM-OCR + Ollama; layout detection, batch processing, and Markdown/DOCX export.
• https://github.com/rav4nn/youtube-rag-scraper — Scrape YouTube channels, extract transcripts, and build FAISS-based semantic search/RAG knowledge bases.
• https://github.com/tonhowtf/omniget — Desktop downloader for 1000+ sites (YouTube, TikTok, Bilibili, Udemy) via yt-dlp; supports torrents and P2P transfers.
• https://github.com/debpalash/VoiceStudio — VoiceStudio (formerly OmniVoice Studio, AGPL-3.0): local voice cloning, voice design, dubbing, and transcription app on the k2-fsa/OmniVoice model; compared with ElevenLabs and V-Clone in Voice Cloning Comparison.
• https://github.com/tronghieuit/valtec-tts — Ultra-lightweight Vietnamese TTS with multi-speaker and zero-shot voice cloning; 74.8M params, CPU-only, 4x realtime.
• https://github.com/jianzongwu/DiffSensei — CVPR 2025 multi-modal LLM + diffusion model for customized black-and-white manga generation; character-aware panel synthesis.
• https://huggingface.co/microsoft/trocr-large-handwritten — Transformer OCR model fine-tuned on IAM for handwritten single text-line recognition.
• https://github.com/nexu-io/html-video — Programmatic HTML-to-video for coding agents; turns HTML, CSS, and data into real MP4s with pluggable render engines, 21 templates, and AI soundtracks (Apache-2.0, no per-render fees).
• https://github.com/nexu-io/html-anything — Agentic HTML editor where your local AI agent writes the HTML; 75 skills across 9 surfaces (magazine, deck, poster, tweet/XHS, prototype, data report), sandboxed preview, and 1-click export to HTML/PNG/WeChat/X/Zhihu with zero API key.
• https://github.com/caotranquochoai/Voice-Translator-V2 — Real-time desktop voice translator (Python/PyQt5); captures mic or system audio and runs an STT → translation → TTS pipeline with multiple swappable engines (Faster Whisper, Qwen3-ASR, Gemini Live, local VieNeu-TTS) for near real-time speech translation.
• https://github.com/kentjuno/ainovel-cli — Multi-agent autonomous long-form novel-writing CLI (Go); a coordinator drives Architect → Writer → Editor in a single LLM loop to go from one prompt to a complete novel (500+ chapters) with 3-tier context management, step-level checkpointing, real-time intervention, and TXT/EPUB export. Vietnamese fork of voocel/ainovel-cli.
• https://novelkit.cc/ — NovelKit, Vietnamese long-form fiction studio. Public code is Novelkit_v2_lite (local FastAPI + React, loopback, one operator, bring-your-own OpenAI-compatible model). License LicenseRef-NovelKit-V2-Lite-NC-ND-1.0: non-commercial, no derivatives, not OSI open source. The homepage describes a Hermes deployment; delegate.py in Lite says that file is a local registry shim for the Hermes delegate_tool. Not NovelAI. See NovelKit.
• https://3dviz.dev/ — Turns a text idea into an interactive 3D scene worth exploring; aimed at creators/designers who want a fast concept-to-3D-visualization workflow.
• https://github.com/altic-dev/FluidVoice — Open-source macOS voice-to-text dictation app (Swift) with on-device AI enhancement; local-first, supports Nemotron/Parakeet/Whisper/Apple speech models, Command Mode and Write Mode, live preview overlay, and optional fully-local "Fluid Intelligence" formatting.
• https://github.com/Zackriya-Solutions/meetily — Privacy-first, self-hosted AI meeting assistant (Rust) for macOS & Windows; 4x-faster Parakeet/Whisper live transcription, speaker diarization, and Ollama-based summarization, 100% local processing with no cloud required.
• https://github.com/wozniakpawel/PairPods — Free, open-source macOS app to share audio between people/devices.
• https://github.com/notepower2k1/CapCap — Windows desktop app for short-form video localization and dubbing, Vietnamese-focused; combines transcription, translation, subtitle styling, voice generation, timeline editing, preview, and export in one project workflow.
• https://github.com/hyscaler/HyCanvas — Free, self-hostable, AI-native design software for creating graphics, social posts, and more, with no paywalls or watermarks.
• https://github.com/krillinai/OpenCreator — Open-source AI workspace for creators (formerly KrillinAI), powered by Codex; agents create videos, images, voice, avatars, translations, and edits all in one place.
• https://github.com/michaelwilhelmsen/humla — Open-source AI meeting notes app for Mac; records mic + system audio with no bot, transcribes on-device or via OpenAI/Deepgram/Groq, identifies speakers offline, writes summaries that fuse your notes with the transcript, and answers questions about your notes with citations (Tauri 2 + Rust + Swift).
• https://vidcdn.one/ — Video Pro CDN: automates video transcoding into multiple qualities (360p–4K), HLS segmentation, and global CDN distribution via distributed cloud worker clusters running 24/7.
• https://ocr.z.ai/ — GLM-OCR hosted service; free OCR that extracts text from images into editable digital text, powered by Zhipu's GLM (the model behind community tools like Folio-OCR above).
• https://huggingface.co/owensong/Inflect-Nano-v1 — Ultra-compact English TTS model (4.63M total inference params including vocoder); full text-to-24kHz-waveform pipeline in a tiny footprint for local/embedded deployment and offline assistants, with expected quality trade-offs at this scale.
• https://ideogram.ai/ — AI image generation platform known for accurate in-image text/typography rendering; text-to-image, product photography, and design tools (logos, posters) alongside more general photorealistic/artistic generation.
• https://sleek.design/ — AI mobile app designer; generates iOS/Android mockups from a text description or reference images, editable visually and exportable to Figma or code (React/HTML).
• https://saydi.ai/ — AI real-time voice translation for meetings/events; 60+ languages, automatic language detection, speaker identification, and integrations with Zoom, Google Meet, and Microsoft Teams.
• https://www.speechsuper.com/ — Deep-learning speech-assessment platform for pronunciation, fluency, grammar, and vocabulary scoring; APIs/SDKs for language-learning apps across 8 languages.
• https://www.pen.dev/ — Agentic design canvas; import content from existing apps, drive edits with multiple AI models (Claude, OpenAI, Gemini, etc.) alongside manual editing, and export designs directly to HTML/CSS.
• https://github.com/open-pencil/open-pencil — OpenPencil: open-source AI-native design editor, Figma alternative. Reads/writes native .fig and .pen files; AI chat with 100+ tools to create/edit nodes (OpenRouter, Anthropic, OpenAI, Google, DeepSeek, Z.ai, MiniMax); headless CLI + XPath queries + the Figma Plugin API via eval + an MCP server for agents (Claude Code, Codex, Gemini CLI); lints naming/layout/accessibility, extracts design tokens, exports selections to JSX/Tailwind; components & variants, image vectorization (Recraft/fal.ai), auto layout + CSS Grid via Yoga WASM; real-time P2P collaboration over WebRTC (no server, no account); a headless Vue SDK for embedding your own editor. Desktop app ~15MB (Tauri v2, macOS/Windows/Linux) or run as a web PWA. MIT.
• https://huggingface.co/briaai/RMBG-2.0 — BRIA AI's background removal / dichotomous image segmentation model (BiRefNet architecture); separates foreground from background and outputs an 8-bit grayscale alpha matte; trained on 15,000+ manually labeled images covering diverse objects and people. SDKs: PyTorch, Transformers, Transformers.js, ONNX. License CC BY-NC 4.0 (non-commercial only; commercial use requires a separate BRIA license or the API).
• https://github.com/mrgoonie/multix-cli — Multi-provider AI multimodal CLI (@mrgoonie/multix, Node.js ≥20): image generation/editing across 8+ providers (Gemini, OpenAI, MiniMax, Leonardo, BytePlus, etc.), video generation and image-to-video, audio (transcription/TTS/music), PDF-to-Markdown conversion, and ffmpeg/ImageMagick-backed media optimization, all behind one unified command pattern instead of juggling separate SDKs. MIT. Also installable as a pi agent extension — see pi-multix in AI Coding Productivity Tools.
• https://synthid.com/ — Google's SynthID Detector, opened to everyone globally in English on 2026-10-07 after a year of limited access for media professionals. Upload an image, video, or audio file to check for the invisible SynthID watermark Google has embedded at generation time since 2023 (180B+ images/videos and 240,000 years of audio watermarked to date). Reads watermarks from Google plus partners OpenAI, NVIDIA, and Kakao, with Apple Intelligence support "soon"; sign-in requires a Google, Apple, or ChatGPT account. Caveat: it only detects SynthID watermarks, so a negative result does not prove the content is human-made, and watermarks can be stripped. The same detection also runs inside Google Search, the Gemini app, and Chrome (1M+ requests/day).

See also: Speech Recognition Tools · Video Generation · Image Generation

• https://github.com/lvt1693/mashloop — Mash Loop: endless local jukebox that finds seamless jump points inside songs and beat-matched, key-locked DJ-style transitions between them (Beat This! + Demucs, Python/FastAPI backend, Web Audio front end); MIT, weeks old, Windows-tested. See Mash Loop.