Skip to content

CRAFT StudioThe Complete YouTube Content Studio

AI scripts. Competitive research. Audio production. GPU-accelerated generation. 14 media sources. One studio.

DocsTypeScriptReactKubernetesAITTSPipelineAgentsGPUMobile

Numbers That Matter โ€‹

14
Media Sources
300+
TTS Voices
14
AI Agents
9
Pipeline Workers

How It Works โ€‹

1
Create a Channel

Define your channel's character, voice, and niche. The AI adapts to your persona.

2
Generate Ideas

AI brainstorms based on your topics, or save inspiration from YouTube research.

3
Run the Pipeline

AI agents research, write, storyboard, and compose your video through 8 stages with automated quality gates.

4
Review & Publish

Producer agent scores each stage. Review the final cut, approve, and publish โ€” all from one dashboard.

Feature Highlights โ€‹

Idea Generation & Management

Generate ideas with AI tuned to your channel's personality. Save inspiration from YouTube Discover. Expand, edit, and convert ideas into full scripts with one click.

  • AI brainstorming with channel context
  • Import from YouTube research with metadata
  • Search, sort, filter by type (shorts/long)
  • Convert to script with enrichment data

AI Script Editor

Write scripts in your channel's character voice. AI-powered revision, polishing, and fact-checking keep your content accurate and natural.

  • Write with Voice โ€” AI drafts in character
  • Revise with custom instructions
  • Fact Check โ€” verify claims with sources
  • Humanize โ€” detect and rewrite AI-sounding text
  • Status workflow: Draft โ†’ Review โ†’ Final

YouTube Discover

Research trending content with yt-dlp-powered search. No API key needed for basic search. Progressive loading fetches more results as you scroll.

  • Filter by duration, date, subscriber cap
  • Outlier detection โ€” find viral hits (5x/10x/50x avg)
  • Channel deep dive with earnings estimates
  • Compare up to 3 channels side-by-side
  • Save any video as an idea with full metadata

Resource Library

14 Sources4 Media TypesAuto Attribution

Search across Pexels, Pixabay, Unsplash, NASA, Wikimedia Commons, Met Museum, Europeana, and more. Preview inline and download with automatic attribution tracking.

  • Video, image, audio, and reference search
  • Grid & list views with hover preview
  • Source and license filtering
  • Tag-based search suggestions from scripts
  • Upload your own media with drag-and-drop

AI Image Generation

Generate thumbnails and B-roll with ComfyUI and Stable Diffusion. Channel-aware AI prompts get you started fast.

  • Text-to-image with multiple checkpoint support
  • Aspect ratio presets (1:1, 16:9, 9:16, 4:3, 3:4)
  • Generation metadata saved per image (prompt, settings)
  • Click-to-zoom preview with full parameter display
  • Per-channel gallery with download and delete

Audio Production

Full audio pipeline from TTS to final mix. 300+ voices with screenplay-aware section parsing. Background music layering and SFX insertion.

  • 4 TTS providers โ€” Edge (free), ElevenLabs, OpenAI, OpenedAI Speech
  • Upload your own voiceover and split at timestamps
  • SFX sections between speech sections
  • Background music with volume, fade, and loop controls
  • RVC voice cloning for custom voice conversion

AI Music Generation

Create instrumental background tracks from text prompts with Meta's MusicGen. Bind directly to audio projects or download standalone.

  • Text-to-music with genre, mood, and instrument control
  • Adjustable duration (10sโ€“30s)
  • AI-assisted prompt generation
  • Attach to audio projects or standalone download
  • Prompt tips for best results

Voice Training (RVC)

Clone any voice with RVC v2. Upload audio samples, browse community models from HuggingFace, and convert TTS output to sound like anyone.

  • Upload voice models (.pth) or training audio (.zip)
  • Browse and install HuggingFace voice models
  • RVC voice conversion on TTS output
  • GPU-accelerated inference
  • Preview voices before applying

Episode Pipeline

8 Stages14 AI AgentsQuality Gates

Orchestrate complete video production through 8 stages โ€” each powered by a specialized AI agent with automated producer review and feedback loops.

  • Research โ†’ Script โ†’ Storyboard โ†’ Assets โ†’ Compositing โ†’ Export โ†’ Review โ†’ Publish
  • Producer agent scores each stage 1-10 with revision feedback
  • Run full pipeline or individual stages
  • Upstream issue detection and soft recovery
  • Real-time progress via SSE

AI Proposals

AI-scored content proposals analyze trends, channel history, and audience signals to surface your best next video topic. Review, approve, and feed directly into the production pipeline.

  • 0-100 confidence scoring with trend signals
  • Timeliness indicators for time-sensitive topics
  • Approve to convert into production-ready ideas
  • Powered by Curator and Trend Analyst agents

Channel Settings & Voice

Configure your channel's personality, voice, and analytics. The AI Character Creator generates a full persona from a few questions.

  • AI Character Creator โ€” tone, quirks, audience
  • Voice picker with 300+ voices and preview
  • Stability and similarity tuning for ElevenLabs
  • RPM presets for earnings estimates
  • MCP server management

Open Source & Self-Hosted โ€‹

Your data stays on your machine. Deploy with Helm on k3s or any Kubernetes cluster โ€” no cloud dependencies beyond your API keys.

bash
helm upgrade --install craft ./helm/craft -f helm/craft/values-dev.yaml

Powered By โ€‹

Frontend
ReactTypeScriptTailwindZustand
Backend
ExpressHelm / k3sPostgreSQLyt-dlpffmpeg
Pipeline
NATS JetStreamAgent SDKRedis9 Workers
AI & Audio
ClaudeGeminiOllamaEdge TTSOpenedAI SpeechMusicGen
GPU Services
ComfyUIStable DiffusionRVC