Claude + HyperFrames Just Solved Video Editing
The creator demonstrates a fully automated video editing pipeline using Claude Code, HyperFrames, and VideoUse to trim raw footage, remove filler words, and add motion graphics — all through natural language prompts. The tutorial walks through setup, tool comparisons, and iterative refinement of AI-generated animations. The core argument is that this workflow replaces manual editing in Adobe Premiere Pro with an orchestrated AI pipeline.
Summary
The video opens with a demonstration of the end result: a 50-second raw clip reduced to 27 seconds with filler words removed, motion graphics added, and subtitles included — all generated by Claude Code without any manual editing. The creator positions this as an evolution of a previous workflow where trimming was still done manually before handing off to HyperFrames for animations.
The creator explains the full tech stack: Claude Code acts as the orchestrator, VideoUse handles transcription and trimming, and HyperFrames renders motion graphics as HTML-based animations. He contrasts HyperFrames with Remotion (VideoUse's native animation tool), showing side-by-side comparisons and arguing that HyperFrames produces more sophisticated, engaging visuals — particularly its 'liquid glass' iOS 26-style UI aesthetic.
The setup process is walked through step by step: installing the Claude Desktop app, choosing a working folder, and pointing Claude at the HyperFrames and VideoUse GitHub repos so it can pull in the necessary skills. The creator also covers API key setup for transcription services, noting that 11Labs, OpenAI Whisper, and a free local tool are all viable options, with 11Labs preferred for cut-point accuracy.
The editing workflow is demonstrated live: a raw 50-second clip is dropped into the project, Claude analyzes the transcript with word-level timestamps, identifies retakes and filler words, and produces a 32-second edited file. The creator then uses voice-to-text to dictate detailed motion graphics instructions — specifying liquid glass cards, karaoke-style text, scissor-cut animations, and a facecam crop transition — and uses Claude's plan mode to review and approve the animation layout before committing tokens to building HTML.
Two rounds of visual iteration are shown: the first pass has issues with the card covering the speaker's face, an unwanted grid overlay, and an incorrect facecam crop. The creator provides specific natural language corrections and the second pass resolves most issues. The HyperFrames timeline editor is highlighted as a tool for quick adjustments without re-prompting Claude.
The creator closes by discussing the long-term vision: building style reference files for different video types (lessons, shorts, intros) so that future videos of the same type can be edited end-to-end by simply dropping in a raw file. He notes the session cost approximately 238,000 tokens and emphasizes that specificity in planning reduces wasted tokens. A screenshot-verification trick — instructing Claude to screenshot rendered frames to self-check output quality — is also shared as a practical tip.
Key Insights
- The creator argues that HyperFrames produces more sophisticated motion graphics than Remotion because it uses HTML-based rendering, resulting in a 'liquid glass' iOS 26-style aesthetic that feels more premium and engaging — even though both tools can handle the animate and render steps of the pipeline.
- The creator explains that word-level timestamp precision from the transcript is critical to the entire pipeline — knowing that a specific word occurs at, say, 11.199 seconds allows motion graphics to be synced exactly to the spoken moment rather than approximated.
- The creator uses a 'teaching a kid to ride a bike' analogy to explain why the first few iterations require very specific directional prompts — the system needs to learn the user's style before it can operate autonomously, and premature hands-off use will produce poor results.
- The creator reveals that the entire live editing session — including planning, building HTML compositions, iterating on visuals, and rendering — consumed approximately 238,000 tokens, and argues that front-loading specificity in the planning stage is the primary lever for reducing token waste.
- The creator describes a self-verification trick where Claude is instructed to take screenshots of rendered frames and check them itself, which causes it to proactively flag visual errors rather than returning a broken output as complete.
Topics
Transcript
[0:00] Claude is now editing my videos end to end. All I have to do is drop in a raw file and it is trimming out the mistakes and the dead space. It's adding motion graphics like you see over here. It's adding dynamic elements like you see over here. It can even add subtitles at the bottom of the screen like you're looking at right here. So, even if you've never coded before or you've never edited videos before, I'm going to show you guys exactly how you can get set up and get outputs that look pretty incredible just like this. All you have to do is use your natural language. It is so insane. Don't want to…
Full transcript available for MurmurCast members
Sign Up to AccessMore from Nate Herk | AI Automation
Fable 5 + Karpathy’s LLM Wiki is Basically Cheating
The video demonstrates how to build a personal LLM-powered knowledge base using Obsidian and Claude, inspired by Andrej Karpathy's approach. By ingesting various data sources (YouTube transcripts, PDFs, URLs) into an interconnected wiki structure, users can create a 'second brain' that helps AI agents understand context and relationships across knowledge domains.
How Claude is Creating a New Generation of Millionaires
Claude, an AI built by Anthropic, is enabling non-technical founders to build million-dollar software companies by automating development work. The transcript outlines why Claude is outpacing competitors, provides real examples like Vulcan, and offers a four-step framework for anyone to start building with Claude today.
How Anthropic Engineers Actually Prompt Fable 5
A guide to effectively prompting Claude Fable 5, covering six key techniques including providing context/intent, using negative prompting, letting the model act once ready, verifying outputs, avoiding reasoning requests, and keeping instructions concise. The speaker emphasizes that Fable 5 is expensive and should only be used 5-15% of the time, with pricing at $10 per million input tokens and $50 per million output tokens.
Stanford's Method Turns Claude Into a PHD Level Research Team
A researcher demonstrates the STORM method from Stanford, which uses five expert perspectives (practitioner, academic, skeptic, economist, historian) to create verified research reports. The method produces 25% more organized articles than competing approaches and is packaged as a reusable Claude skill that generates HTML briefings with peer-reviewed citations.
Is Claude Mythos Coming?
A YouTuber analyzes the brief appearance of 'Claude Mythos' on Anthropic's API, arguing it signals a marketing move rather than an imminent public launch. Despite competitive pressure from OpenAI and IPO timing, the creator believes Mythos capabilities will quietly fold into future Opus models rather than release publicly under that name. Three possible scenarios are outlined ranging from a limited gated release to Mythos remaining permanently restricted to vetted security partners.