Claude Code can now make music and sound effects for free, on your own computer. Audio Studio MCP connects Claude to Stable Audio 3 running locally. Ask in plain words, like "cozy lo-fi cafe music" or "glass bottle shattering on concrete", and it: → generates 1 to 4 versions per request → restyles a clip, redoes part of it, or extends a track → makes seamless loops, fades and even loudness → exports 48 kHz WAV, OGG or FLAC, ready for games and videos No cloud, no API key, and generating is free. You need an Apple Silicon Mac (16 GB RAM recommended) or a Windows PC with an NVIDIA GPU (8 GB+ VRAM). Install on Mac: git clone cd audio-studio-mcp ./install.sh The installer prints the command to connect it to Claude Code. Check the Stable Audio 3 model license before using sounds commercially. Listen to samples:

Hermes is getting a LOCAL video editor. 🪽 Not “generate me a video.” Edit MY footage. Cut. Join. Crop. Captions. Overlays. Audio sync. Loudness. Reels/TikTok checks. 42 FFmpeg editing scripts. No cloud. No API key. This demo is going to be ridiculous. 🔥

Livestream Alert: Using Comfy to Go Beyond the Models: Custom Workflows for Commercial and Film Production Join us for a conversation with @eyefyah (Creative Director) and Ariel Klevecz (Tech Lead/Content Automation) from Tool, on using ComfyUI as a flexible production tool rather than a single model or plugin. Instead of centering on any one model, Ezra and Ariel will walk through how Tool builds custom Comfy workflows around what a specific project actually needs, and how those workflows get integrated into traditional filmmaking, VFX, and post-production pipelines. What we'll cover: Building custom Comfy workflows around what a production actually needs, not what a model does by default Precise first-person hand/performance control and extreme-resolution outpainting with Wan 2.1 (AMD CES film) Audio-reactive generation and letting a song's loudness constrain what Wan paints, frame-chained (Flower) Comfy in a commercial pipeline generating a different world per execution (Starburst) Comfy-generated compositing layers, stop-motion, and title sequences moving between Comfy, After Effects, and Flame (Everywhere/Nowhere)

🌙 Tonight's thesis — picks frozen Monday Jul 27 The index frame: SPX band 7,371–7,421, lean CALLS (skew +0.08) — the tape earns a bullish tilt. Fed-liquidity read: neutral (z -0.4, 20d -0.0). Standing back from: Hyperscalers (-4.3%/5d), Software (-2.4%/5d), Power / DC Infra (-2.2%/5d) — longs there are discounted until the 5-day trend itself recovers (a one-day pause is a dead-cat, the worst state in our backtest). The top of the book, and why: - CVX CALL (risk-adj vote +1.4) — Net edge here — running into CVX's upcoming catalyst (buy-the-rumor), though Energy is soft. (also: MACD) - MCD PUT (risk-adj vote -1.1) — Net edge here — fading MCD's momentum (though Defensive is bid today). (also: Wall Position) - NFLX CALL (risk-adj vote +1.1) — Net edge here — running into NFLX's upcoming catalyst (buy-the-rumor), as money rotates into Growth Tech. (also: 3-Day Up Streak) - BRK-B CALL (risk-adj vote +1.0) — Net edge here — running into BRK-B's upcoming catalyst (buy-the-rumor). (also: Pre-Event Run-up) Risk posture: 33 calls / 35 puts screened (puts carry a ×1.25 evidence bar — the tape drifts up). Every vote here is risk-adjusted (σ+β), liquidity-taxed, factor-capped at 3 per correlated complex, and stripped of earnings coin-flips. High conviction means diverse agreement per unit of risk, not loudness. The trade: buy ~3:50 PM ET, sell ~9:40 AM ET. 📋 The full board — 33 bet-up / 35 bet-down, ranked in the card 🐲 $SPX #Stocks# #OptionsTrading#

System prompt Create a complete animated short film from this brief. Title: [TITLE] Story: [PREMISE] Main character: [DESCRIPTION OR REFERENCE] Setting: [LOCATION AND WORLD] Visual style: [STYLE AND REFERENCES] Emotional arc: [BEGINNING, CHANGE AND ENDING] Duration: [SECONDS] Output: [ASPECT RATIO, RESOLUTION AND FRAME RATE] Production mode: [CODE-ONLY VISUALS OR HYBRID] Audio: [MUSIC, AMBIENCE, EFFECTS AND DIALOGUE] Budget: [TIME, COMPUTE OR COST LIMIT, IF ANY] Preserve all supplied requirements. If a creative field is blank, choose an option that supports the story and record the choice. If output settings are blank, use 16:9, 1920 × 1080 and 24 fps. Use the requested settings when they are provided. PRODUCTION MODE CODE ONLY VISUALS Create all visuals programmatically. Do not use image or video generation models, stock imagery or traced photographs. Visual assets rendered or cached by the project code are allowed. Audio is a separate choice. Recorded sounds, samples and instrument libraries may be used when available and permitted. Record their sources. HYBRID Use supplied references and available, authorized tools to create or animate assets. Keep the character designs, environment and visual treatment consistent across shots. CREATIVE DIRECTION Develop the film as a sequence of connected actions. Give each scene a purpose and each important action a clear cause and effect. Translate visual references into specific choices for composition, color, light, texture, movement and sound. Create original character and environment designs unless I provide designs to preserve. Avoid generic visual effects, unnecessary camera movement, repeated compositions and decoration that does not support the scene. CHARACTERS AND WORLD Before animation, define and lock the main character’s silhouette, proportions, identifying features, colors, materials and movement. If a reference is provided, preserve its recognizable features. Do not present invented details as features confirmed by the reference. Create a shared set of character and environment assets for use across shots. Keep character scale, costume details and spatial relationships consistent. Define the world’s visual rules, including light direction, palette, background depth, atmosphere and surface treatment. Change these deliberately when the story or setting requires it. STORY AND SHOTS Create a timed shot plan covering the complete film. For each shot, specify: - Frame range and duration. - Story purpose and emotional beat. - Subject position, action and timing. - Framing, camera angle and movement. - Environment, lighting and color. - Transition and sound. Express cut points as whole frames. Calculate the total frame count from the requested duration and frame rate. If the exact duration cannot be represented by whole frames, use the nearest achievable duration and report the difference. Do not add camera movement without a visual or narrative purpose. Hold the camera still when the action benefits from it. ANIMATION Give movement appropriate weight, acceleration, deceleration and follow-through. Make movement originate from the subject’s action. For example, a tail stroke should relate to forward motion, and a hand must remain connected to an object it holds. Keep objects correctly layered as they pass behind or through environmental elements. Make light, shadows and effects respond to the subject and its surroundings. Use simulation or approximation where it serves the shot. Identify approximations accurately in the production notes. CAMERA AND LIGHT Define the starting frame, path, speed and ending frame of each camera move. Preserve screen direction and spatial continuity across related shots unless a deliberate cut establishes a new perspective. Make environmental light affect both the subject and background. Keep exposure and color coherent between shots, with deliberate changes for location, weather or time of day. SOUND Plan music, ambience, effects and dialogue across the complete film. Synchronize sounds with visible actions. Use level, tone and space to communicate distance. Leave room for silence when the scene calls for it. Choose a loudness target suited to the intended delivery platform. Record the target and measured result. Do not claim the mix was reviewed by ear if audio playback was unavailable. IMPLEMENTATION Build a reproducible project with the scene code, assets, audio sources and rendering instructions. Select tools appropriate to the brief and available environment. Record important dependency versions. Use a controlled timeline based on frame number and frame rate. Seed procedural variation. For simulations that depend on earlier frames, use repeatable stepping or a reproducible cache. Render and encode the full film at the requested settings. Do not silently reduce the final resolution, frame rate or duration. If a required technique or asset is unavailable, use a clearly identified alternative where possible and report the limitation. PRODUCTION ORDER 1. Define the story, shot plan and visual rules. 2. Lock the character and environment designs. 3. Create representative style frames. 4. Test the most demanding action in a short animated shot. 5. Review and fix the test. 6. Produce and assemble the complete film. 7. Review the integrated film and improve the largest problems. 8. Render and verify the final export. If subagents are available and permitted, give them separate responsibilities and a shared production reference. Check their work together for continuity. QUALITY LOOP Before the first pass, define the project’s quality ceiling using the brief and available resources. Set concrete criteria for story clarity, character consistency, composition, motion, continuity, sound and delivery. Do not set an iteration count in advance. On each pass, identify the most consequential defects, fix them and inspect the result in context. Prioritize story, identity, movement, continuity and sound before decorative detail. Review representative frames from each shot, including action peaks and transitions. Watch the film at normal speed when playback is available. Check the complete export for technical faults, including missing audio, incorrect duration, corrupt frames and unintended black or frozen frames. Exclude intentional silence, holds and black frames. A contact sheet does not establish motion quality. An automated check does not establish artistic quality. Report any review that could not be performed. Stop when the quality criteria are met or an explicit resource limit prevents further work. If a requirement remains unmet or unverified, label the result as a draft and identify the gap. DELIVERABLES Provide: 1. The final video, or a clearly labeled draft. 2. The editable project and assets needed to reproduce it. 3. The timed shot list. 4. Character and visual references. 5. Representative contact sheets. 6. Verified export metadata and render settings. 7. Audio sources and mix notes. 8. A concise list of completed checks and remaining gaps. Report only work, tests and reviews that were actually completed.

The loudest voices stoking fears about AI dangers have made tremendous headway in the past two weeks. AI technology has not taken some unexpected, dangerous turn, but the hype around it — propelled by what appears to be a well orchestrated PR campaign — has drummed up considerable fear. I worry that it represents a setback for our field. I have written frequently that fears of AI are overhyped. AI’s capabilities can be uncannily human-like and unpredictable, and it’s rational to worry when people who are directly involved express concerns. But I see the problems as a sign of the engineering work that ahead, rather than insurmountable barriers or the sky falling. AI technology continues to advance — which is a good thing! — but technical advances, poorly understood by the public, give those who seek to generate hype repeated opportunities to do so. First, I don’t see any step up in the risk of human extinction from AI compared to a few months ago. The theories about this remain the same fantastical, science fiction scenarios as a few months ago. The biggest change in AI risk is its cybersecurity capabilities — a topic which we should take seriously — but this, too, will not lead to the end of the world. The most notable recent event leading to increased fear was when an OpenAI team deployed an agent swarm that hacked into Hugging Face. Much of the popular press contained significant hype. For example, some publications reported that a swarm of 1,200 agents carried out the attack. While this was technically accurate, as I write this, I have about 1,300 processes running on my laptop. Yes, the ability to get large swarms of agents to work in parallel on a task is a significant technical advance, And, in computing, many processes run at the same time. So this shouldn’t be seen as some magical capability. Additionally, OpenAI’s buggy sandboxing and monitoring processes were key to enabling this incident. Fixing these bugs and putting in place improved monitoring would be appropriate fixes, not pausing AI. There are many well known ways to attack software systems. The main advantage of AI agents is that they are relentless. They will tirelessly try many tactics — and have the patience to chain vulnerabilities together — that previously would have taken an infeasible amount of human effort. But in the long term, I believe the advantage will lie with defenders (because they have more information with which to identify bugs, which they can fix), but the cyber-threat landscape has changed significantly. There are still bottlenecks to identifying and exploiting a vulnerability. AI agents still have to try a lot of things to see what works, and taking these actions takes time and might be detected by defenders. This is why, even though it is now easy to obtain versions of leading open weight models that have had their guardrails removed or weakened, so they will not refuse to try to execute cyber attacks, the world has not ended. I am also concerned about the anthropomorphization of AI in a lot of reporting, where LLMs and agents are unnecessarily treated as if they were people. If I wield a hammer, miss a nail, and accidentally dent the wall, it’s not the fault of the hammer. The problem lies in how I used the hammer. Similarly, if I prompt an agent and it hacks into someone else’s system, the responsibility lies with me, not the agent. Of course, we want to build systems that are as safe and predictable as possible. (For example, an unsafe hammer would be one whose head randomly flies off under normal use.) Today’s agentic systems are not predictable, but I see no reason why, by applying sound engineering practices, we won’t be able to make them extremely safe to use. One new element in the forecasts of AI-enabled doom is AI companies disclaiming responsibility for their own products. “I didn’t do it; my out-of-control agent did!” There’s a balance to be struck between the responsibility of the tool maker and the tool user, but when something goes wrong, let’s hold the people building and/or using the hammer responsible, rather than the hammer. (By the way, if you’re worried about AI bioweapon risk, David Bellamy has a great post on why this, too, is overhyped. Briefly, the bottleneck in building a bioweapon is not intelligence, but lab work and manufacturing.) Pausing AI progress will create much more harm than benefit. First, our adversaries will certainly not slow down. Second, engineering requires discovering problems empirically so we can fix them. If we pause AI by a decade, we will also delay finding and implementing safety engineering fixes by about the same duration. Of course, the incentive to stoke fears — for regulatory capture, to garner attention, or to make one’s technology seem more powerful — remains the same as before. Disclaiming responsibility is a new one. Taking a hard technical look at the actual risks however, I see little factual basis for the degree of fear that’s been stoked up. We still have hard research and engineering work ahead to improve AI safety, but the beneficial applications continue to vastly outweigh the risks, and we should keep building. [Original text (with links): ]