LLMs and AI tools — useful links, organized

A practical, curated directory for fast access to the AI services and learning resources that are most useful in practice: chats, APIs, model hubs, local LLM tools, benchmarks, courses, builders to follow, video, image, voice, coding, research and agents. Curated rather than exhaustive.

Last reviewed: September 2026 · External links open in a new tab.

General-purpose AI / LLM chat

Model hubs and local LLMs

  • Hugging Face Modelscore — models, GGUFs, datasets and Spaces
  • Ollama — simple local model runtime
  • LM Studio — desktop local LLM UI and server
  • llama.cppcore — efficient GGUF inference
  • Unsloth — model fine-tuning and quantized releases
  • vLLM — high-throughput serving
  • Open WebUI — self-hosted chat interface
  • Jan — local/open-source desktop assistant

Learn AI — structured resources

People to follow — AI startups & business

Avatars, dubbing and video localization

  • HeyGenfeatured — avatars, voice cloning, lip-sync and video translation
  • Synthesia — enterprise avatar video
  • Captions — AI avatars, editing and social video
  • D-ID — talking avatars and agents

Open-source voice cloning — free/local

  • Qwen3-TTS Basebest overallApache-2.0 — 0.6B/1.7B models, multilingual, voice cloning from a short reference; official Gradio demo
  • Chatterbox MultilingualeasyMIT — strong zero-shot cloning, emotion control, pip install + Gradio UI
  • GPT-SoVITSeasy WindowsMIT — 5-second zero-shot TTS, WebUI and downloadable Windows package
  • OpenVoice V2MIT — lightweight instant voice cloning with multilingual support
  • CosyVoice 3Apache-2.0 — high-quality multilingual and cross-lingual zero-shot cloning; more technical setup
  • F5-TTSofficial weights: non-commercial — excellent reference-audio TTS and Gradio UI; code is MIT but official pretrained weights are CC-BY-NC
  • Fish Speech / Fish Audio S2research / non-commercial weights — very strong multilingual TTS and voice generation; check model-license terms

Fastest starting points: try Qwen3-TTS or Chatterbox for a modern local alternative to ElevenLabs; GPT-SoVITS is especially convenient on Windows. Clone only voices you own or have permission to use.

Speech-to-text and transcription

Research and productivity

  • NotebookLMcore — source-grounded notebooks
  • Perplexity — web research with citations
  • Gamma — presentations and documents
  • Canva AI — design and presentation tools
  • Notion AI — workspace search and writing
  • Genspark — research and agentic workflows

Agents and automation

  • Manusagent
  • Google Opal — no-code AI mini-apps and workflows
  • Genspark — general-purpose agent workflows
  • n8n — automation with AI nodes and agents
  • Zapier AI — app automation and agents
  • Make — visual workflow automation

Selection rule: prioritize tools and learning sources that are broadly useful, technically relevant or good gateways to an ecosystem. For fast-moving model names, links point to durable product, repository or platform pages whenever possible.