Voice & Multimodal — AI Agents, Skills & Tools
Agents, skills, guides, tools, and commands for voice & multimodal — 57 curated resources for building with AI coding agents.
Voice Agent Engineer
Use this agent to build or fix a real-time voice agent — the streaming STT → LLM → TTS pipeline, conversational (mouth-to-ear) latency, turn-taking, barge-in/interruptions, and per-stage provider selection. Examples — "our voice bot feels laggy and talks over people, fix the turn-taking and latency", "build a phone agent that transcribes, answers with our LLM, and speaks back", "get our voice agent's response time under a second".
Design Systems Librarian
Use this agent to keep a design system's tokens, components, and documentation consistent with each other — finding hardcoded colors, spacing, and radii that should be tokens, components with no docs and docs describing variants that no longer exist, naming drift between token names, component names, and design-tool variables, and stale documentation — then reporting the drift with severity and a prioritized fix list. Examples — 'audit our design system for hardcoded values', 'which components are undocumented or documented wrong?', 'our token names and the Figma variables have diverged, show me where'.
Multimodal Document Extractor
Extract structured data from documents and images with a vision-language model — define the target schema, prompt the VLM to fill it from the page (invoices, forms, receipts, statements, IDs), and verify critical fields against the source. Use when you need reliable structured output from messy, varied, or scanned documents that defeat template-based OCR.
Component Spec Writer
Turn a component screenshot or written description into a developer handoff spec covering anatomy, variants, states, a props table with types and defaults, behavior, responsive rules, edge cases, accessibility notes for an auditor to verify, and open questions, keeping what the artifact shows separate from what was inferred and turning every gap into a question instead of a plausible default. Use when a component is going to engineering and the design file is the only documentation that exists.
Design Brief Writer
Turn a messy design request you paste in (a Slack thread, an email chain, meeting notes, a ticket) into a one-page design brief: goal, audience, constraints, success criteria, scope and non-goals, references needed, open questions, and timeline placeholders, with every item the source did not state marked as an inference and no constraint or deadline invented. Use when a request has arrived in fragments and the brief has to exist before design work starts.
Design Critique Checklist
Critique an attached screenshot or a described screen against a fixed seven-part checklist (visual hierarchy, spacing and alignment, typography, color and contrast as observed, empty, loading, and error states, consistency, and copy) and return findings in one set report format, each with a location, a severity, and a suggested fix, deferring every WCAG measurement to an accessibility audit. Use when you want a repeatable design review of a screen that reads the same from one round to the next.
Design Token Extractor
Read a screenshot, a pasted stylesheet, or a written component description and produce a design token set as W3C DTCG-style JSON, with colors, a type scale, spacing, radii, and shadows grouped and named, followed by a notes list saying which values were read exactly, snapped to a scale, or estimated by eye. Use when a design exists only as pixels or ad-hoc CSS and you need a named token set to start from.
UX Copy Reviewer
Review a set of interface strings you paste in (buttons, labels, empty states, errors, confirmations, tooltips) as one set rather than one at a time, checking clarity, term consistency across screens, error and empty-state quality, whether each action label matches what happens, and length against any limits you supply, and returning a verdict and a rewrite per string without inventing a product fact the strings do not contain. Use when microcopy has accumulated across a product and nobody has read it end to end.
The Best AI Image Generators in 2026 (for Design Work)
AI image generators compared for design work in 2026: text rendering, vector output, style control, commercial licensing, API access, and pricing model.
The Best AI Tools for Designers in 2026
The AI tools designers actually use in 2026: prototyping, websites, image generation, decks, design-to-code, and design systems, with a verdict per category.
The Best Document Parsers for RAG in 2026: LlamaParse vs Docling vs Unstructured vs Reducto
LlamaParse, Docling, Unstructured, Reducto, Marker and Mistral OCR compared on deployment, license, OCR, tables and output, with a verdict for each.
Claude Design vs Figma Make: Which Prompt-to-Prototype Tool?
Claude Design vs Figma Make compared on what each makes, plan and seat requirements, design-system input, refinement, export, and handoff to code.
Claude Code for Designers: Prototype in Code Without Becoming an Engineer
Why a designer opens a terminal agent, how to set it up safely, connecting Figma, a CLAUDE.md for a design-system repo, and three workflows with prompts.
Claude Design: The Complete Guide (2026)
What Claude Design makes, how it ingests your design system, the three editing modes, every export and handoff, and where it stops. Written for designers.
Anthropic's Design Plugin for Claude: Every Skill Explained
Every skill in Anthropic's open-source design plugin for Claude Cowork and Claude Code, the connectors it expects, the install commands, and what it leaves out.
Claude Skills for Designers: 5 to Upload Today
Five Claude skills for designers: brief, critique, token extraction, component spec, and UX copy review. What each returns, a prompt for each, and upload steps.
Figma to Code with Claude Code and the Figma MCP Server
Connect Figma's MCP server to Claude Code and turn a selected frame into a component that uses your real design tokens and your existing components.
Maintaining a Design System with Claude Code
Tokens, components, docs, and the Figma file drift apart within weeks. A repo layout, CLAUDE.md rules, and three Claude Code jobs that catch it every week.
Which Claude Plan Should a Designer Pay For?
A decision guide by situation: solo prototyping, a team sharing one design system, heavy Design and Cowork use, or an agency. Plan names, no prices.
Add Image Understanding to Your App
A practical guide to sending images to a vision model and getting reliable, structured results: base64 vs URL, resolution, prompting, cost, and errors.
Multimodal Embeddings and Image Search
How multimodal embeddings put images and text in one vector space, and how to build text-to-image and image-to-image search on top of it.
Multimodal RAG over PDFs, Scans & Charts: Two Approaches That Actually Work
RAG over visual documents — PDFs, scans, charts — where text-only extraction loses tables and layout. Parse-then-text vs embed-the-page-image, with trade-offs.
Screenshot-to-Code: Building UIs from Images with AI
Turn a screenshot, mockup, or Figma frame into working frontend code with AI vision models — the realistic workflow, the right tools, and the honest pitfalls.
Vision-Language Models Compared (2026)
Which vision-language model to reach for, by job: Claude, GPT, Gemini, and open models like Qwen3-VL compared on OCR, charts, grounding, video, and cost.
Using Vision-Language Models for OCR, Documents, and Video Understanding
How to use vision-language models for OCR, documents, and video: how they differ from traditional OCR, their failure modes, and getting reliable output.
Best Speech-to-Text APIs in 2026
The STT field, honestly ranked — Deepgram and AssemblyAI's hosted duel, Whisper as the open baseline, Cartesia Ink for latency — and how to pick by workload.
Best Text-to-Speech APIs in 2026
The TTS APIs worth building on — ElevenLabs for quality and breadth, Cartesia Sonic for realtime latency — and how to choose for agents vs produced audio.
How to Build a Voice Agent: The STT → LLM → TTS Pipeline
How to build a real-time voice agent: the STT → LLM → TTS pipeline, the latency budget that makes or breaks it, and how to wire each stage.
Realtime Voice Agents: Build on LiveKit, Buy Vapi, or Pipeline with Pipecat
The three ways to ship a realtime voice agent in 2026 — open infrastructure, managed platform, or OSS pipeline — and how speech-to-speech models fit in.
Assemblyai
Speech AI platform: promptable Universal-3 Pro STT, a flat-rate Voice Agent API, and speech understanding — summarization, sentiment, PII redaction.
Canva
A design platform whose Magic Studio and Canva AI features generate designs, copy, and edits on top of your brand kit, across web, desktop, and mobile.
Cartesia
Real-time voice AI on state-space models — Sonic streaming TTS, Ink STT with native turn detection, and Line, a code-first voice-agent platform.
Claude Design
An Anthropic Labs tool for making prototypes, decks, and one-pagers with Claude, using your design system, with PPTX, PDF, HTML, and Canva export.
Deepgram
A voice-AI platform with fast, accurate speech-to-text (Nova) and low-latency text-to-speech (Aura), plus a bundled Voice Agent API.
Descript
A video and podcast editor where you edit by editing the transcript, with Underlord, an AI co-editor, Studio Sound, AI voices, dubbing, and clip creation.
ElevenLabs
A voice-AI platform for high-quality text-to-speech, voice cloning, dubbing, and real-time conversational agents, via API.
fal
fal is a generative-media inference cloud for running image, video, and audio diffusion models fast — 1,000+ models, a simple API, and pay-per-use pricing.
Figma Make
Figma's prompt-to-app tool: describe an idea or paste a design and get a working, code-backed prototype you can keep editing inside Figma.
Framer AI
A website design tool with AI agents that generate editable pages, wireframes, CMS content, and code components, then publish to hosting Framer runs.
Ideogram
An image generator built around reliable text rendering, with prompt tooling, background control, on-brand custom models, an API, and an MCP server.
Livekit
Open-source realtime infrastructure — a WebRTC server plus the LiveKit Agents framework for production voice AI, with turn detection, telephony, and cloud.
Midjourney
An image and video generator with a web app and a Discord bot, known for strong aesthetic defaults, style references, an inpainting editor, and paid-only plans.
Opus Clip
An AI clipping tool that turns long videos into scored, captioned short clips for Shorts, TikTok, and Reels, with a scheduler, brand templates, and an API.
Pipecat
An open-source Python framework for real-time voice and multimodal conversational AI — it orchestrates streaming STT, LLM, and TTS into composable pipelines.
Qwen3-VL
Alibaba Qwen's open-weights vision-language model family (2B–235B, Apache-2.0): image and document understanding, OCR, visual reasoning, and video.
Recraft
An AI image platform built for designers, with editable vector output, reusable custom styles, mockups, and an API for image and vector generation.
Relume
An AI site builder that turns a brief into a sitemap, wireframes, and a designed marketing site, then exports to Figma, Webflow, or React.
Stitch
Google's AI design canvas that turns prompts, voice, images, or an existing codebase into high-fidelity mobile and web UI, with frontend code and a Figma paste.
Vapi
The API-first voice-agent platform — assemble phone and web agents from any STT/LLM/TTS mix, with telephony, squads, and tool calling handled for you.
Whisper
OpenAI's open-weights speech-to-text — the MIT-licensed multilingual model family that made self-hosted transcription a default, with a huge ecosystem.
Critique Screen
Read a UI screenshot from a path or URL, run the fixed design critique checklist over it, and write the report to design/critiques/ without touching the design or the code.
Design Tokens
Extract design tokens from a stylesheet, a Tailwind or theme config, or a screenshot, then create or update design/tokens.json and report every added, changed, and removed token.
Design-to-Code
Design-to-code is turning a design — a Figma file, a screenshot, or a generated mockup — into working front-end code, increasingly by an AI agent.
Design Tokens
Design tokens are named values for a design system's decisions — color, spacing, type, radius — stored once and referenced everywhere instead of hardcoded.
Diffusion Model
A diffusion model generates an image by starting from random noise and removing it step by step, guided by a prompt, until a coherent picture remains.
Generative UI
Generative UI is an interface produced by a model from a prompt — a whole screen, flow, or prototype generated rather than assembled by hand.
Text-to-Image
Text-to-image is generating a picture from a written prompt, using a model trained to turn a description into pixels — or, in some tools, vectors.