Turn massive volumes of audio and video
into structured professional deliverables.
8 tabs. One pipeline. AI at every step.
An AI Guide helps you plan the perfect workflow. A Custom Prompt Engine designs the perfect extraction. AI Curation scores and sorts everything. You approve.
Paste URLs from 8 platforms. Import entire channels, playlists, or paste 100+ URLs at once. XTC downloads the audio and queues everything for processing.
Videos, channels, playlists. Set range bounds.
Reels, stories, IGTV.
Any public TikTok URL.
Spaces, video, voice tweets.
Public and unlisted videos.
Podcast episodes.
RSS feeds, direct episode URLs.
MP3, MP4, WAV, M4A, more.
Four engines from fast drafts to studio-grade accuracy. AI-powered speaker diarization separates voices automatically. 90+ languages supported.
Whisper tiny/base. Fastest turnaround.
Whisper medium/large. Balanced.
Whisper large-v3 with diarization.
Cloud engine. Native diarization.
AI-powered transcript cleanup. Fix transcription errors, rename speakers, remove filler words, and merge segments. Integrated waveform editor for precise fixes when you need to verify against the audio.
Remove the noise before you extract the signal. Strip intros, ads, outros, filler segments, and anything else that dilutes your source material.
Semantic clustering groups hundreds of extractions into topics automatically. No manual sorting. Related quotes, insights, and dialogues land together.
Three providers. Your API keys. Pick the model that fits your budget and quality needs.
Best for nuanced extraction and long-form content. Excellent at preserving voice and tone.
Reliable extraction with strong instruction following. Great all-rounder.
Fast, capable models from xAI. Growing model lineup.
Keyboard-driven curation built for speed. Review extractions, keep the diamonds, discard the rest. Process hundreds of results in minutes, not hours.
This is where XTC goes from extraction tool to book-writing pipeline. AI drafts full book chapters from your curated selections. Hierarchical outline editor, side-by-side preview, and real-time generation.
Export your work in six formats. Whether you need raw data for a pipeline, polished chapters for a publisher, or structured notes for your CMS.
25+ built-in extraction prompts across 5 categories. Write your own. Refine until it is right. Save as presets.
Theme Quotes, Punchy/Short, Quote+Context, Actionable, Contrarian/Insight
Best Exchanges, Q&A Highlights, Conflict/Tension
Rewrite spoken content as polished, publishable prose
Strip speaker labels and timecodes for clean output
Your saved custom prompts, reusable across sessions
Write your own extraction logic. Iterate until output matches your needs.
Run an extraction, review the output, adjust your prompt, run again. Iterate until the results are exactly what you need. No guessing.
See the estimated LLM cost before running any extraction. Strict fidelity toggle keeps extractions close to the original wording when accuracy matters most.
Seven AI-powered capabilities that turn XTC Studio from a pipeline into a thinking partner. Every feature runs locally, works offline, and respects your data.
Describe what you need in plain language. The AI Guide interviews you, understands your goal, and plans the entire pipeline: which sources to pull, which transcription engine to use, what extraction prompts to run, and how to organize the output.
No need to learn the 8-tab pipeline. Tell the Guide what you want, and it configures everything for you.
Skip the manual setup on complex projects. Describe the deliverable, review the plan, hit go.
Two fundamentally different ways to process content. Preserve mode keeps exact words for books and legal work. Distill mode extracts ideas, themes, and insights for research and course creation.
Exact words. Faithful to the source. Every quote, every phrase, every nuance stays intact. Built for authors compiling books from spoken material, lawyers needing verbatim testimony, and journalists who need the real words.
Extract the signal, not the noise. AI identifies core ideas, synthesizes themes across sources, and produces clean, structured insights. Built for researchers, course creators, and anyone turning hours of content into knowledge.
AI scores every extracted block from 1 to 5 and recommends Keep, Star, or Trash. It explains its reasoning so you stay in control. Style-aware: it curates differently for a legal brief than for a research paper.
Prioritizes verbatim statements, dates, names, and legally significant language. Flags hearsay and speculation.
Surfaces clinical findings, dosages, patient outcomes, and diagnostic reasoning. Filters small talk.
Weights novel insights, data points, methodology, and citations. Deprioritizes repetition and filler.
Tell the AI what you're building. It scans your content, understands the structure, and designs a precision extraction prompt tailored to your exact deliverable. No prompt engineering required.
You describe the output. The engine reverse-engineers the extraction logic. No trial and error, no wasted API calls.
The engine reads a sample of your transcripts before generating the prompt. It adapts to your specific content, not generic templates.
When you process 50+ sources, the same idea shows up in different words across different files. Duplicate Detection finds blocks saying the same thing, clusters them, and lets you pick the best version. Runs locally. Free.
One click sends your curated content straight to your Obsidian vault. Wikilinks, frontmatter, tags, and folder structure included. Your extractions become a connected knowledge graph, not a pile of files.
Toggle on source timestamps for every extracted block. Filmmakers and editors can jump straight to the exact moment in the original recording. Researchers can verify quotes against source material in seconds.
Find the exact take. Click the timecode, hear the moment, mark it for your edit.
Verify every quote against the original audio. No more scrubbing through hours of footage.
Cite with precision. Every extracted insight links back to the exact moment it was spoken.
macOS 13 (Ventura) or later.
Dependencies: Python 3.10+, Homebrew, conda. All auto-installed on first launch.
Join the waitlist for early access.