[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M77T0E64C26C0MNENCAYSC0E":3,"skill-same:01M77T0E64C26C0MNENCAYSC0E":71},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":25,"commitSha":26,"skillDirectory":27,"forksCount":28,"repositoryHtmlUrl":29,"repositoryHomepageUrl":6,"ownerHtmlUrl":30,"ownerBlogUrl":31,"ownerDescription":6,"updatedAt":32,"install":33,"resources":46},"01M77T0E64C26C0MNENCAYSC0E","hyperframes-audio",null,"Use when audio already placed in a HyperFrames composition needs to be mixed: a music bed that fights a voiceover (voiceover carve), effects on a track (EQ, compressor, limiter, gate, saturation, delay, reverb, chorus, phaser, bitcrush), or automation envelopes drawn on a track's volume or any effect parameter. Don't use for sourcing or generating audio — finding BGM, SFX, or making a voiceover is `\u002Fmedia-use`. Don't use for clip timing or track layout, which is `\u002Fhyperframes-core`.","Reihantt6\u002Fsns-myagent","Reihantt6","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F219822095?v=4","01M77KK94PV8B7C2D9XWXNS8DJ","Mohamad Reihan tegar prasetya","01M77KK958QFMTHYXGXFV7KZ79",2,"MIT","https:\u002F\u002Fgithub.com\u002FReihantt6\u002Fsns-myagent\u002Fblob\u002Fmain\u002Fagent\u002Fskills\u002Fhyperframes-audio\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\ndescription: \"Use when audio already placed in a HyperFrames composition needs to be mixed: a music bed that fights a voiceover (voiceover carve), effects on a track (EQ, compressor, limiter, gate, saturation, delay, reverb, chorus, phaser, bitcrush), or automation envelopes drawn on a track's volume or any effect parameter. Don't use for sourcing or generating audio — finding BGM, SFX, or making a voiceover is `\u002Fmedia-use`. Don't use for clip timing or track layout, which is `\u002Fhyperframes-core`.\\n\"\n---\n# HyperFrames Audio\n\nA mix is a set of relationships, not a stack of processors. Two tracks that each\nsound right alone can be unlistenable together, and the fix is almost never \"turn\none down\" — it is finding what they are fighting over and giving it to whichever\none needs it. Every tool here exists to express one of those relationships.\n\nEffects live on the element as `data-fx-chain`, and preview and render run the\nsame Web Audio graph — the studio in a live context, the engine in an offline one\ninside the browser it already drives. There is one implementation of each effect,\nso what you hear while scrubbing is what gets written. You never tune twice.\n\nThree attributes carry everything, all on the audio\u002Fvideo element itself:\n\n| Attribute         | Holds                                                     |\n| ----------------- | --------------------------------------------------------- |\n| `data-fx-chain`   | the effects, in signal order                              |\n| `data-automation` | envelopes on this track's volume or its effect parameters |\n| `data-fx-carve`   | the carve's own settings, so it can be re-derived         |\n\nExact JSON for each, and the rules a lane must satisfy: `references\u002Fattributes.md`.\nEvery effect with its parameters, ranges and units: `references\u002Ffx-registry.md`.\nHow to work out what is wrong with a file you cannot hear:\n`references\u002Fdiagnosis.md`.\n**Presets, named jobs and one-knob profiles, plus a symptom-to-fix table:\n`references\u002Fpresets.md`** — read that before hand-building a chain, because one\nof the presets or named jobs usually already names the problem.\n\n## How it fits together\n\nTwo authoring surfaces write those attributes; two runtimes read them through the\nsame builders. That shared middle is why preview predicts the render.\n\n```mermaid\nflowchart TB\n  voice[\"voice track\u003Cbr\u002F>media file\"]\n  bed[\"music bed\u003Cbr\u002F>media file\"]\n\n  subgraph AUTHOR[\"Authoring — the only things that write attributes\"]\n    panel[\"Studio\u003Cbr\u002F>Voiceover carve control\"]\n    script[\"scripts\u002Fcarve.mjs\u003Cbr\u002F>detects the pair, dynamic by default\"]\n    analysis[\"core\u002FaudioCarve.ts\u003Cbr\u002F>carveProfile · analyseCarveBands\u003Cbr\u002F>analyseCarveDuck · analyseCarveDynamics\"]\n    panel --> analysis\n    script --> analysis\n  end\n\n  voice --> analysis\n  bed --> analysis\n\n  subgraph ATTRS[\"Written onto the bed element\"]\n    carveAttr[\"data-fx-carve\u003Cbr\u002F>source · strength · dynamic\"]\n    chainAttr[\"data-fx-chain\u003Cbr\u002F>peaking xN + gain, tagged fromCarve\"]\n    autoAttr[\"data-automation\u003Cbr\u002F>a lane per carved parameter\"]\n  end\n\n  analysis --> carveAttr\n  analysis --> chainAttr\n  analysis --> autoAttr\n\n  subgraph SHARED[\"One implementation, read by both\"]\n    build[\"audioFxGraph.ts · buildFxChain\"]\n    sched[\"audioFxAutomation.ts · scheduleChainAutomation\"]\n  end\n\n  chainAttr --> build\n  autoAttr --> sched\n\n  build --> preview[\"Preview\u003Cbr\u002F>live AudioContext\u003Cbr\u002F>attachElementFxChain\"]\n  sched --> preview\n  build --> render[\"Render\u003Cbr\u002F>OfflineAudioContext in the headless browser\u003Cbr\u002F>applyAudioFxChain\"]\n  sched --> render\n\n  preview --> heard[\"what you hear while scrubbing\"]\n  render --> wav[\"processed WAV\u003Cbr\u002F>+ chainTailSeconds so the mix lets the tail through\"]\n  wav --> mix[\"engine · audioMixer\u003Cbr\u002F>volume lane baked into the PCM here, not in the graph\"]\n  mix --> out[\"the rendered mix\"]\n\n  edit[\"editing the attribute mid-playback\"] -.->|MutationObserver| preview\n```\n\nThe carve's own settings are never read at playback — the chain and lanes it\nproduced are what play. `data-fx-carve` exists so strength can be changed on an\nexisting carve instead of guessed back out of the filters.\n\nInside a carved bed the signal runs through the dips first, then the level match,\nthen anything you built yourself — which is why a limiter you add still acts as\nthe last ceiling:\n\n```mermaid\nflowchart LR\n  src[\"decoded bed\"] --> p1[\"peaking\u003Cbr\u002F>400 Hz\"]\n  p1 --> p2[\"peaking\u003Cbr\u002F>1 kHz\"]\n  p2 --> p3[\"peaking\u003Cbr\u002F>1.6 kHz\"]\n  p3 --> g[\"gain\u003Cbr\u002F>level match\"]\n  g --> hand[\"your own effects\u003Cbr\u002F>e.g. limiter\"]\n  hand --> dest[\"track gain, then out\"]\n\n  l1[\"lane fx.n1.gain\"] -.->|\"envelope of the voice's\u003Cbr\u002F>level in that band\"| p1\n  l4[\"lane fx.n4.gain\"] -.->|\"how far the bed\u003Cbr\u002F>ducks overall\"| g\n```\n\nA static carve is the same graph with fixed values and no lanes at all.\n\n## First, work out what is wrong\n\nThe table below starts from \"it sounds boomy\" — which presumes somebody already\nlistened and said so. Handed a file and \"fix this\", you have no such sentence\nand you cannot listen, so you have to measure. One rule governs all of it:\n\n> **The absolute spectrum of a single unknown voice cannot be diagnosed.**\n> Formants are ±10 dB, fundamentals run 85–255 Hz, and sentences decline 5–6 dB\n> as they end. Every one of those reads as a defect on its own, and every one of\n> them is the speaker.\n\nSo compare, and compare against something **inside the same file**: the clean\noriginal if it exists, otherwise the pauses — whatever is audible in a gap is\nadditive, and the gap's spectrum is the channel rather than the voice. Comparing\nagainst a published average spectrum or a synthesised control voice does not\nwork: two speakers differ by more than most defects, and both wrong answers in\nthe evaluation behind this guidance came from exactly that.\n\nWhen there is no original and no usable silence, a static tonal defect is\ngenuinely under-determined. Say so and offer the readings that fit, rather than\npicking one and building a chain on it.\n\nCommands, traps and worked recipes: **`references\u002Fdiagnosis.md`**. Read it\nbefore diagnosing a file nobody has described.\n\n## Start from the symptom\n\nOnce you know the band and the kind, name what is wrong with the audio. Most bad audio is\none or two of these, and each has a shipped answer:\n\n| It sounds like                     | Reach for                                          |\n| ---------------------------------- | -------------------------------------------------- |\n| Hum or thump underneath            | `rumble-cut`, or a `highpass` at 80 Hz             |\n| Boomy, chesty                      | **Tame Boominess** job (200 Hz)                    |\n| Muffled, behind cardboard          | **Reduce Mud** job (250 Hz)                        |\n| Words hard to make out             | **Add Clarity** job (3 kHz), or carve the bed      |\n| Harsh and tiring                   | **Soften Harshness** job (3.2 kHz)                 |\n| Some words much louder than others | **Evenness** on a compressor, or Even Out Levels   |\n| Room tone between sentences        | `room-gate`                                        |\n| Voice and music fighting           | **Voiceover carve** — not an EQ on either          |\n| Dry, recorded nowhere              | `room-tight` or `room-natural`                     |\n| Just \"amateur\"                     | `voice-clean`, which is four of the above in order |\n\nFull catalogue, what each preset contains, the band vocabulary, and what is\ndeliberately NOT covered (de-essing, noise removal, tone match):\n`references\u002Fpresets.md`.\n\nSubtract before you add, level after you filter, relationships after level,\ncharacter and ceiling last. Each step changes what the next one hears — a\ncompressor set before a high-pass spends its time chasing rumble.\n\n## Reach for a family by the problem, not the name\n\n**Filters** (`highpass`, `lowpass`, `peaking`, `lowshelf`, `highshelf`) decide\nwhich frequencies a track is allowed to occupy. This is the first tool for two\nsources colliding, because collisions happen in bands: a bed and a voice both\nwant 1–3 kHz, and taking that from the bed costs the bed far less than turning\nthe whole thing down costs the mix. A high-pass on a voice is the standard fix\nfor rumble; a low-pass darkens or muffles deliberately.\n\n**Dynamics** (`gain`, `compressor`, `limiter`, `gate`) decide how a track's level\nbehaves over time. Compression narrows the distance between loud and quiet so the\nquiet parts can come up. A limiter is a ceiling — it does not shape anything, it\nguarantees nothing gets past. A gate removes what is below a threshold, which is\nhow you silence room tone between phrases. `gain` is a plain level stage, and it\nis what an automation lane rides when a track has to move out of the way.\n\n**Nonlinear** (`saturate`, `bitcrush`) changes the waveform's shape, which adds\nharmonics that were not there. Reach for it when a track needs character or\ngrit rather than correction — and remember it is generative: it makes a thin\nsource denser, not cleaner.\n\n**Time** (`delay`, `reverb`, `chorus`, `phaser`) puts a track in a space or gives\nit width. These are the ones that most easily wreck a mix, because a tail or a\ndetuned copy occupies the same room a voice needs. Use them on the thing that\nshould sit _behind_ something else, and keep the wet amount lower than sounds\nright in isolation.\n\nThe chain is serial: each effect processes what the one before it produced. So\ncorrective filtering goes early, character in the middle, and a limiter last\nwhere it can actually act as a ceiling.\n\n## Voiceover carve\n\n**The problem it solves.** A music bed under a voice makes the voice hard to\nfollow. The reflex is to duck the whole bed, which works and costs the bed all of\nits presence — the music goes limp for the entire voiceover. But the voice does\nnot need the whole spectrum. It needs the few bands it actually occupies. Carve\ntakes only those, and the bed keeps its low end and its top, so it is still music\nwhile the voice is still intelligible.\n\n**It is a relationship, not an effect.** The settings live on the _bed_ — the\ntrack that gets processed — and they name the voices to listen to, exactly as a\nsidechain compressor does: you select the track that gets quieter and pick what\nmakes it quieter. **Never put a carve on a voice track.** A voice carved against\nitself is a bug, not a subtle mix choice.\n\n**Every voice, not one of them.** `sources` is a list, because a bed usually runs\nunder a whole sequence — a narrator, an interview answer, a second presenter. They\nare summed onto the bed's own clock before anything is measured (`mixCarveSources`),\nso one analysis covers all of them: the bands come from all the speech there is, and\nthe envelopes rise wherever any of it is happening. Voices that never play while the\nbed does are left out; they cannot mask it.\n\n**One knob.** `strength` is 0..1 and derives everything: how deep to cut, how\nmany bands, how wide, how far to favour intelligibility over raw voice energy,\nhow far the level may drop, how far under the voice to aim. Those six move\ntogether in any real mix — a gentle carve is a shallow cut in few bands with\nlittle ducking, a hard one is deeper in more bands with more — so they are one\nrelationship written once, in `carveProfile`. Default is `0.25` — a 6 dB dip in\nthree bands with 6 dB of level room, audible without sounding like a hole. At\n`0.5` the dip reaches 10 dB, which is where a carve starts being heard as an\neffect rather than as room for the voice; above that is deliberate territory for\na loud bed under a quiet voice. `0` is spectral only — one band, no level match\nat all.\n\n**Carve by default.** A bed playing under narration wants a carve; it is not a\npolish step to get to if there is time. Place both tracks, run the command below,\nlisten. Skip it only when there is no narration for the music to sit under — a\nmusic video, a title card, a montage cut to the track.\n\n**It always follows the voice.** There is no static mode: a fixed depth thins the\nbed through every pause, and once you have heard both there is no reason to want it.\nEvery value becomes an envelope of the speech's own level — silence leaves the bed\nalone, a loud passage pushes the carve to full depth — written as ordinary automation,\nwhich is why the lanes show up in the timeline and can be edited afterwards.\n\n**Level matching is part of it.** Spectral carving cannot fix a bed that is\nsimply louder than the voice. So the carve also measures how far over the voice\nthe bed sits and writes a `gain` stage: held at one value for a static carve,\ndriven by an envelope for a dynamic one. That envelope releases slowly on\npurpose — music that snaps back to full the instant a word ends sounds like a\nmachine doing it.\n\n**Running it.** In Studio the carve is one module at the top of a track's effect\nrack — voice, strength, dynamic, and the analysis it produced, in one card. It is\nthere whenever another track could be the voice, and a bed with exactly **one**\ncandidate above it is carved by default, dynamically, at the default strength:\nthat is what a bed under narration wants, and the module is where you change or\nswitch it off. Several candidates leaves the picker waiting rather than guessing.\nHeadless —\nwhich is the path when you are authoring a composition rather than editing one:\n\n```bash\nnode \u003CSKILL_DIR>\u002Fscripts\u002Fcarve.mjs --comp index.html\n```\n\nThat is the whole command. It finds the voice and the bed itself, carves\ndynamically at the default strength, and prints what it decided:\n\n```\nbed    music-bed (name looks like music)\nvoice  narration (only track left)\ncarve  strength 0.25 dynamic\nbands  400Hz -6dB q1.4, 1000Hz -3dB q1.4, 1600Hz -3.17dB q1.4\nlevel  216-point envelope, floor -6 dB\n```\n\nName the tracks with `--bed` \u002F `--voice` (repeatable) when the automatic choice is\nwrong, `--strength` to push it, `--dry-run` to see that report and write nothing.\n\n**How it picks the tracks.** Names first, because that is what you already told it\nand the answer is explainable — `classifyAudioName` in core, the same classifier\nStudio's own picker uses, so the two cannot disagree. A track whose id or filename\nlooks like music (`music`, `bgm`, `bed`, `score`…) is the bed; everything else that\nplays over it and is not SFX-shaped is a voice. Audio elements are preferred: video\ncounts only when no audio track is left to be the voice, or every B-roll clip in the\ncomposition would read as somebody talking. **It refuses when it cannot tell which\ntrack is the bed** rather than carving the wrong one — typing one id is cheap.\n\nSame analysis functions as the panel, so the result is identical. Needs `ffmpeg`\non PATH and `@hyperframes\u002Fcore` installed in the project (`npm i -D\n@hyperframes\u002Fcore`) — the CLI inlines core rather than shipping it, so it cannot\nbe borrowed from there.\n\n**What it writes** is an ordinary chain of peaking filters plus a gain stage,\ntagged `fromCarve`. That tagging is the whole trick: a re-run replaces the\nprevious carve and leaves every effect you built by hand — and every lane you\ndrew by hand — exactly where it was. So re-carving at a new strength is safe and\nrepeatable, and `data-fx-carve` exists so the settings can be read back rather\nthan guessed from the filters.\n\n## Automation\n\nA lane is a set of breakpoints on one parameter: `{t, v}` in clip-local seconds\nand the parameter's own units. Targets are `volume` for the track's level, or\n`fx.\u003CnodeId>.\u003Cparam>` for an effect's knob.\n\n**Only some parameters can be automated, and a lane on the others is silently\ninert.** A knob is automatable when a Web Audio `AudioParam` backs it. The four\nworklet-based effects — `compressor`, `limiter`, `gate`, `bitcrush` — expose\nnone at all, so no lane on any of their parameters will ever move: to make a\ncompressor's behaviour change over time, automate a `gain` stage before it\ninstead. `references\u002Ffx-registry.md` marks every parameter.\n\n## Verify\n\nAlmost no static gate covers the mix. The linter reads `data-automation` for\nexactly one conflict — `audio_volume_double_automation`, a volume lane on a track\nthat also has a GSAP tween on `volume`, where the lane wins and the tween is\nignored — and nothing validates the chain or the effect lanes at all. What\nenforces those is the render: a chain it cannot parse fails the whole mix rather\nthan quietly writing the dry signal, because a mix that sounds plausible and is\nwrong is worse than a refusal. Preview is the opposite by design: an unreadable\nchain plays dry so the composition stays workable.\n\nA lane pointing at a node the chain does not have is pruned on read, not an\nerror — so a typo'd `nodeId` costs you the envelope silently. Read the ids back\nout of the chain rather than assuming what was minted.\n\nEffects with a tail (`reverb`, `delay`) make the rendered track **longer** than\nits source, and the mix is told how much by the chain. So a bed with reverb no\nlonger ends exactly at its `data-duration`; that is expected, not a bug.\n\nBeyond that, a mix is verified by rendering and listening. For a carve: the voice\nshould be legible without the bed sounding hollowed, and with `dynamic` the bed\nshould come back up between phrases rather than staying flat. If the bed sounds\nnotched rather than simply quieter under the voice, the strength is too high —\nthat is the one failure mode with an obvious sound.\n","# HyperFrames Audio\n\nA mix is a set of relationships, not a stack of processors. Two tracks that each\nsound right alone can be unlistenable together, and the fix is almost never \"turn\none down\" — it is finding what they are fighting over and giving it to whichever\none needs it. Every tool here exists to express one of those relationships.\n\nEffects live on the element as `data-fx-chain`, and preview and render run the\nsame Web Audio graph — the studio in a live context, the engine in an offline one\ninside the browser it already drives. There is one implementation of each effect,\nso what you hear while scrubbing is what gets written. You never tune twice.\n\nThree attributes carry everything, all on the audio\u002Fvideo element itself:\n\n| Attribute         | Holds                                                     |\n| ----------------- | --------------------------------------------------------- |\n| `data-fx-chain`   | the effects, in signal order                              |\n| `data-automation` | envelopes on this track's volume or its effect parameters |\n| `data-fx-carve`   | the carve's own settings, so it can be re-derived         |\n\nExact JSON for each, and the rules a lane must satisfy: `references\u002Fattributes.md`.\nEvery effect with its parameters, ranges and units: `references\u002Ffx-registry.md`.\nHow to work out what is wrong with a file you cannot hear:\n`references\u002Fdiagnosis.md`.\n**Presets, named jobs and one-knob profiles, plus a symptom-to-fix table:\n`references\u002Fpresets.md`** — read that before hand-building a chain, because one\nof the presets or named jobs usually already names the problem.\n\n## How it fits together\n\nTwo authoring surfaces write those attributes; two runtimes read them through the\nsame builders. That shared middle is why preview predicts the render.\n\n```mermaid\nflowchart TB\n  voice[\"voice track\u003Cbr\u002F>media file\"]\n  bed[\"music bed\u003Cbr\u002F>media file\"]\n\n  subgraph AUTHOR[\"Authoring — the only things that write attributes\"]\n    panel[\"Studio\u003Cbr\u002F>Voiceover carve control\"]\n    script[\"scripts\u002Fcarve.mjs\u003Cbr\u002F>detects the pair, dynamic by default\"]\n    analysis[\"core\u002FaudioCarve.ts\u003Cbr\u002F>carveProfile · analyseCarveBands\u003Cbr\u002F>analyseCarveDuck · analyseCarveDynamics\"]\n    panel --> analysis\n    script --> analysis\n  end\n\n  voice --> analysis\n  bed --> analysis\n\n  subgraph ATTRS[\"Written onto the bed element\"]\n    carveAttr[\"data-fx-carve\u003Cbr\u002F>source · strength · dynamic\"]\n    chainAttr[\"data-fx-chain\u003Cbr\u002F>peaking xN + gain, tagged fromCarve\"]\n    autoAttr[\"data-automation\u003Cbr\u002F>a lane per carved parameter\"]\n  end\n\n  analysis --> carveAttr\n  analysis --> chainAttr\n  analysis --> autoAttr\n\n  subgraph SHARED[\"One implementation, read by both\"]\n    build[\"audioFxGraph.ts · buildFxChain\"]\n    sched[\"audioFxAutomation.ts · scheduleChainAutomation\"]\n  end\n\n  chainAttr --> build\n  autoAttr --> sched\n\n  build --> preview[\"Preview\u003Cbr\u002F>live AudioContext\u003Cbr\u002F>attachElementFxChain\"]\n  sched --> preview\n  build --> render[\"Render\u003Cbr\u002F>OfflineAudioContext in the headless browser\u003Cbr\u002F>applyAudioFxChain\"]\n  sched --> render\n\n  preview --> heard[\"what you hear while scrubbing\"]\n  render --> wav[\"processed WAV\u003Cbr\u002F>+ chainTailSeconds so the mix lets the tail through\"]\n  wav --> mix[\"engine · audioMixer\u003Cbr\u002F>volume lane baked into the PCM here, not in the graph\"]\n  mix --> out[\"the rendered mix\"]\n\n  edit[\"editing the attribute mid-playback\"] -.->|MutationObserver| preview\n```\n\nThe carve's own settings are never read at playback — the chain and lanes it\nproduced are what play. `data-fx-carve` exists so strength can be changed on an\nexisting carve instead of guessed back out of the filters.\n\nInside a carved bed the signal runs through the dips first, then the level match,\nthen anything you built yourself — which is why a limiter you add still acts as\nthe last ceiling:\n\n```mermaid\nflowchart LR\n  src[\"decoded bed\"] --> p1[\"peaking\u003Cbr\u002F>400 Hz\"]\n  p1 --> p2[\"peaking\u003Cbr\u002F>1 kHz\"]\n  p2 --> p3[\"peaking\u003Cbr\u002F>1.6 kHz\"]\n  p3 --> g[\"gain\u003Cbr\u002F>level match\"]\n  g --> hand[\"your own effects\u003Cbr\u002F>e.g. limiter\"]\n  hand --> dest[\"track gain, then out\"]\n\n  l1[\"lane fx.n1.gain\"] -.->|\"envelope of the voice's\u003Cbr\u002F>level in that band\"| p1\n  l4[\"lane fx.n4.gain\"] -.->|\"how far the bed\u003Cbr\u002F>ducks overall\"| g\n```\n\nA static carve is the same graph with fixed values and no lanes at all.\n\n## First, work out what is wrong\n\nThe table below starts from \"it sounds boomy\" — which presumes somebody already\nlistened and said so. Handed a file and \"fix this\", you have no such sentence\nand you cannot listen, so you have to measure. One rule governs all of it:\n\n> **The absolute spectrum of a single unknown voice cannot be diagnosed.**\n> Formants are ±10 dB, fundamentals run 85–255 Hz, and sentences decline 5–6 dB\n> as they end. Every one of those reads as a defect on its own, and every one of\n> them is the speaker.\n\nSo compare, and compare against something **inside the same file**: the clean\noriginal if it exists, otherwise the pauses — whatever is audible in a gap is\nadditive, and the gap's spectrum is the channel rather than the voice. Comparing\nagainst a published average spectrum or a synthesised control voice does not\nwork: two speakers differ by more than most defects, and both wrong answers in\nthe evaluation behind this guidance came from exactly that.\n\nWhen there is no original and no usable silence, a static tonal defect is\ngenuinely under-determined. Say so and offer the readings that fit, rather than\npicking one and building a chain on it.\n\nCommands, traps and worked recipes: **`references\u002Fdiagnosis.md`**. Read it\nbefore diagnosing a file nobody has described.\n\n## Start from the symptom\n\nOnce you know the band and the kind, name what is wrong with the audio. Most bad audio is\none or two of these, and each has a shipped answer:\n\n| It sounds like                     | Reach for                                          |\n| ---------------------------------- | -------------------------------------------------- |\n| Hum or thump underneath            | `rumble-cut`, or a `highpass` at 80 Hz             |\n| Boomy, chesty                      | **Tame Boominess** job (200 Hz)                    |\n| Muffled, behind cardboard          | **Reduce Mud** job (250 Hz)                        |\n| Words hard to make out             | **Add Clarity** job (3 kHz), or carve the bed      |\n| Harsh and tiring                   | **Soften Harshness** job (3.2 kHz)                 |\n| Some words much louder than others | **Evenness** on a compressor, or Even Out Levels   |\n| Room tone between sentences        | `room-gate`                                        |\n| Voice and music fighting           | **Voiceover carve** — not an EQ on either          |\n| Dry, recorded nowhere              | `room-tight` or `room-natural`                     |\n| Just \"amateur\"                     | `voice-clean`, which is four of the above in order |\n\nFull catalogue, what each preset contains, the band vocabulary, and what is\ndeliberately NOT covered (de-essing, noise removal, tone match):\n`references\u002Fpresets.md`.\n\nSubtract before you add, level after you filter, relationships after level,\ncharacter and ceiling last. Each step changes what the next one hears — a\ncompressor set before a high-pass spends its time chasing rumble.\n\n## Reach for a family by the problem, not the name\n\n**Filters** (`highpass`, `lowpass`, `peaking`, `lowshelf`, `highshelf`) decide\nwhich frequencies a track is allowed to occupy. This is the first tool for two\nsources colliding, because collisions happen in bands: a bed and a voice both\nwant 1–3 kHz, and taking that from the bed costs the bed far less than turning\nthe whole thing down costs the mix. A high-pass on a voice is the standard fix\nfor rumble; a low-pass darkens or muffles deliberately.\n\n**Dynamics** (`gain`, `compressor`, `limiter`, `gate`) decide how a track's level\nbehaves over time. Compression narrows the distance between loud and quiet so the\nquiet parts can come up. A limiter is a ceiling — it does not shape anything, it\nguarantees nothing gets past. A gate removes what is below a threshold, which is\nhow you silence room tone between phrases. `gain` is a plain level stage, and it\nis what an automation lane rides when a track has to move out of the way.\n\n**Nonlinear** (`saturate`, `bitcrush`) changes the waveform's shape, which adds\nharmonics that were not there. Reach for it when a track needs character or\ngrit rather than correction — and remember it is generative: it makes a thin\nsource denser, not cleaner.\n\n**Time** (`delay`, `reverb`, `chorus`, `phaser`) puts a track in a space or gives\nit width. These are the ones that most easily wreck a mix, because a tail or a\ndetuned copy occupies the same room a voice needs. Use them on the thing that\nshould sit _behind_ something else, and keep the wet amount lower than sounds\nright in isolation.\n\nThe chain is serial: each effect processes what the one before it produced. So\ncorrective filtering goes early, character in the middle, and a limiter last\nwhere it can actually act as a ceiling.\n\n## Voiceover carve\n\n**The problem it solves.** A music bed under a voice makes the voice hard to\nfollow. The reflex is to duck the whole bed, which works and costs the bed all of\nits presence — the music goes limp for the entire voiceover. But the voice does\nnot need the whole spectrum. It needs the few bands it actually occupies. Carve\ntakes only those, and the bed keeps its low end and its top, so it is still music\nwhile the voice is still intelligible.\n\n**It is a relationship, not an effect.** The settings live on the _bed_ — the\ntrack that gets processed — and they name the voices to listen to, exactly as a\nsidechain compressor does: you select the track that gets quieter and pick what\nmakes it quieter. **Never put a carve on a voice track.** A voice carved against\nitself is a bug, not a subtle mix choice.\n\n**Every voice, not one of them.** `sources` is a list, because a bed usually runs\nunder a whole sequence — a narrator, an interview answer, a second presenter. They\nare summed onto the bed's own clock before anything is measured (`mixCarveSources`),\nso one analysis covers all of them: the bands come from all the speech there is, and\nthe envelopes rise wherever any of it is happening. Voices that never play while the\nbed does are left out; they cannot mask it.\n\n**One knob.** `strength` is 0..1 and derives everything: how deep to cut, how\nmany bands, how wide, how far to favour intelligibility over raw voice energy,\nhow far the level may drop, how far under the voice to aim. Those six move\ntogether in any real mix — a gentle carve is a shallow cut in few bands with\nlittle ducking, a hard one is deeper in more bands with more — so they are one\nrelationship written once, in `carveProfile`. Default is `0.25` — a 6 dB dip in\nthree bands with 6 dB of level room, audible without sounding like a hole. At\n`0.5` the dip reaches 10 dB, which is where a carve starts being heard as an\neffect rather than as room for the voice; above that is deliberate territory for\na loud bed under a quiet voice. `0` is spectral only — one band, no level match\nat all.\n\n**Carve by default.** A bed playing under narration wants a carve; it is not a\npolish step to get to if there is time. Place both tracks, run the command below,\nlisten. Skip it only when there is no narration for the music to sit under — a\nmusic video, a title card, a montage cut to the track.\n\n**It always follows the voice.** There is no static mode: a fixed depth thins the\nbed through every pause, and once you have heard both there is no reason to want it.\nEvery value becomes an envelope of the speech's own level — silence leaves the bed\nalone, a loud passage pushes the carve to full depth — written as ordinary automation,\nwhich is why the lanes show up in the timeline and can be edited afterwards.\n\n**Level matching is part of it.** Spectral carving cannot fix a bed that is\nsimply louder than the voice. So the carve also measures how far over the voice\nthe bed sits and writes a `gain` stage: held at one value for a static carve,\ndriven by an envelope for a dynamic one. That envelope releases slowly on\npurpose — music that snaps back to full the instant a word ends sounds like a\nmachine doing it.\n\n**Running it.** In Studio the carve is one module at the top of a track's effect\nrack — voice, strength, dynamic, and the analysis it produced, in one card. It is\nthere whenever another track could be the voice, and a bed with exactly **one**\ncandidate above it is carved by default, dynamically, at the default strength:\nthat is what a bed under narration wants, and the module is where you change or\nswitch it off. Several candidates leaves the picker waiting rather than guessing.\nHeadless —\nwhich is the path when you are authoring a composition rather than editing one:\n\n```bash\nnode \u003CSKILL_DIR>\u002Fscripts\u002Fcarve.mjs --comp index.html\n```\n\nThat is the whole command. It finds the voice and the bed itself, carves\ndynamically at the default strength, and prints what it decided:\n\n```\nbed    music-bed (name looks like music)\nvoice  narration (only track left)\ncarve  strength 0.25 dynamic\nbands  400Hz -6dB q1.4, 1000Hz -3dB q1.4, 1600Hz -3.17dB q1.4\nlevel  216-point envelope, floor -6 dB\n```\n\nName the tracks with `--bed` \u002F `--voice` (repeatable) when the automatic choice is\nwrong, `--strength` to push it, `--dry-run` to see that report and write nothing.\n\n**How it picks the tracks.** Names first, because that is what you already told it\nand the answer is explainable — `classifyAudioName` in core, the same classifier\nStudio's own picker uses, so the two cannot disagree. A track whose id or filename\nlooks like music (`music`, `bgm`, `bed`, `score`…) is the bed; everything else that\nplays over it and is not SFX-shaped is a voice. Audio elements are preferred: video\ncounts only when no audio track is left to be the voice, or every B-roll clip in the\ncomposition would read as somebody talking. **It refuses when it cannot tell which\ntrack is the bed** rather than carving the wrong one — typing one id is cheap.\n\nSame analysis functions as the panel, so the result is identical. Needs `ffmpeg`\non PATH and `@hyperframes\u002Fcore` installed in the project (`npm i -D\n@hyperframes\u002Fcore`) — the CLI inlines core rather than shipping it, so it cannot\nbe borrowed from there.\n\n**What it writes** is an ordinary chain of peaking filters plus a gain stage,\ntagged `fromCarve`. That tagging is the whole trick: a re-run replaces the\nprevious carve and leaves every effect you built by hand — and every lane you\ndrew by hand — exactly where it was. So re-carving at a new strength is safe and\nrepeatable, and `data-fx-carve` exists so the settings can be read back rather\nthan guessed from the filters.\n\n## Automation\n\nA lane is a set of breakpoints on one parameter: `{t, v}` in clip-local seconds\nand the parameter's own units. Targets are `volume` for the track's level, or\n`fx.\u003CnodeId>.\u003Cparam>` for an effect's knob.\n\n**Only some parameters can be automated, and a lane on the others is silently\ninert.** A knob is automatable when a Web Audio `AudioParam` backs it. The four\nworklet-based effects — `compressor`, `limiter`, `gate`, `bitcrush` — expose\nnone at all, so no lane on any of their parameters will ever move: to make a\ncompressor's behaviour change over time, automate a `gain` stage before it\ninstead. `references\u002Ffx-registry.md` marks every parameter.\n\n## Verify\n\nAlmost no static gate covers the mix. The linter reads `data-automation` for\nexactly one conflict — `audio_volume_double_automation`, a volume lane on a track\nthat also has a GSAP tween on `volume`, where the lane wins and the tween is\nignored — and nothing validates the chain or the effect lanes at all. What\nenforces those is the render: a chain it cannot parse fails the whole mix rather\nthan quietly writing the dry signal, because a mix that sounds plausible and is\nwrong is worse than a refusal. Preview is the opposite by design: an unreadable\nchain plays dry so the composition stays workable.\n\nA lane pointing at a node the chain does not have is pruned on read, not an\nerror — so a typo'd `nodeId` costs you the envelope silently. Read the ids back\nout of the chain rather than assuming what was minted.\n\nEffects with a tail (`reverb`, `delay`) make the rendered track **longer** than\nits source, and the mix is told how much by the chain. So a bed with reverb no\nlonger ends exactly at its `data-duration`; that is expected, not a bug.\n\nBeyond that, a mix is verified by rendering and listening. For a carve: the voice\nshould be legible without the bed sounding hollowed, and with `dynamic` the bed\nshould come back up between phrases rather than staying flat. If the bed sounds\nnotched rather than simply quieter under the voice, the strength is too high —\nthat is the one failure mode with an obvious sound.\n",{"description":24},"Use when audio already placed in a HyperFrames composition needs to be mixed: a music bed that fights a voiceover (voiceover carve), effects on a track (EQ, compressor, limiter, gate, saturation, delay, reverb, chorus, phaser, bitcrush), or automation envelopes drawn on a track's volume or any effect parameter. Don't use for sourcing or generating audio — finding BGM, SFX, or making a voiceover is `\u002Fmedia-use`. Don't use for clip timing or track layout, which is `\u002Fhyperframes-core`.\n","main","58843b159d7a4050d6b785a98b908cce505c0b92","agent\u002Fskills\u002Fhyperframes-audio",0,"https:\u002F\u002Fgithub.com\u002FReihantt6\u002Fsns-myagent","https:\u002F\u002Fgithub.com\u002FReihantt6","reihsns.web.id","2026-08-19 20:08:18",{"source":8,"skillName":5,"commands":34,"prompt":44,"zipAvailable":45},[35,38,41],{"tool":36,"command":37},"npx","npx skills add Reihantt6\u002Fsns-myagent --skill hyperframes-audio",{"tool":39,"command":40},"bunx","bunx skills add Reihantt6\u002Fsns-myagent --skill hyperframes-audio",{"tool":42,"command":43},"pnpm","pnpm dlx skills add Reihantt6\u002Fsns-myagent --skill hyperframes-audio","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FReihantt6\u002Fsns-myagent\u002Fblob\u002Fmain\u002Fagent\u002Fskills\u002Fhyperframes-audio\u002FSKILL.md",true,[47,53,57,61,65],{"path":48,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":51,"sourceHtmlUrl":52},"references\u002Fattributes.md","REFERENCE","参考文档",5905,"https:\u002F\u002Fgithub.com\u002FReihantt6\u002Fsns-myagent\u002Fblob\u002Fmain\u002Fagent\u002Fskills\u002Fhyperframes-audio\u002Freferences\u002Fattributes.md",{"path":54,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":55,"sourceHtmlUrl":56},"references\u002Fdiagnosis.md",10519,"https:\u002F\u002Fgithub.com\u002FReihantt6\u002Fsns-myagent\u002Fblob\u002Fmain\u002Fagent\u002Fskills\u002Fhyperframes-audio\u002Freferences\u002Fdiagnosis.md",{"path":58,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":59,"sourceHtmlUrl":60},"references\u002Ffx-registry.md",6928,"https:\u002F\u002Fgithub.com\u002FReihantt6\u002Fsns-myagent\u002Fblob\u002Fmain\u002Fagent\u002Fskills\u002Fhyperframes-audio\u002Freferences\u002Ffx-registry.md",{"path":62,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":63,"sourceHtmlUrl":64},"references\u002Fpresets.md",13155,"https:\u002F\u002Fgithub.com\u002FReihantt6\u002Fsns-myagent\u002Fblob\u002Fmain\u002Fagent\u002Fskills\u002Fhyperframes-audio\u002Freferences\u002Fpresets.md",{"path":66,"resourceType":67,"resourceTypeDesc":68,"sizeBytes":69,"sourceHtmlUrl":70},"scripts\u002Fcarve.mjs","SCRIPT","脚本",15319,"https:\u002F\u002Fgithub.com\u002FReihantt6\u002Fsns-myagent\u002Fblob\u002Fmain\u002Fagent\u002Fskills\u002Fhyperframes-audio\u002Fscripts\u002Fcarve.mjs",[72,79,86,93,100,107],{"skillNo":73,"name":74,"nameZh":6,"description":75,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":76,"categories":77,"keywords":78,"publishedAt":19},"01M77XDCYZ407XYMF2T4P6F7VT","youtube-transcript","Extract transcripts from YouTube videos. Use when the user asks for a transcript, subtitles, or captions of a YouTube video and provides a YouTube URL (youtube.com\u002Fwatch?v=, youtu.be\u002F, or similar). Supports output with or without timestamps.","https:\u002F\u002Fgithub.com\u002FReihantt6\u002Fsns-myagent\u002Fblob\u002Fmain\u002Fagent\u002Fskills\u002Fyoutube-transcript\u002FSKILL.md",[],[],{"skillNo":80,"name":81,"nameZh":6,"description":82,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":83,"categories":84,"keywords":85,"publishedAt":19},"01M77XD7PBC98J6MXGJJ2TJ8DX","youtube-thumbnail-design","YouTube thumbnail design with specific dimensions, contrast rules, and mobile preview optimization. Covers safe zones, text placement, face expression psychology, and A\u002FB testing. Use for: YouTube thumbnails, video cover images, click-through optimization. Triggers: youtube thumbnail, thumbnail design, video thumbnail, click through rate, ctr optimization, youtube cover, video cover image, thumbnail maker, thumbnail tips, youtube design, video preview image","https:\u002F\u002Fgithub.com\u002FReihantt6\u002Fsns-myagent\u002Fblob\u002Fmain\u002Fagent\u002Fskills\u002Fyoutube-thumbnail-design\u002FSKILL.md",[],[],{"skillNo":87,"name":88,"nameZh":6,"description":89,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":90,"categories":91,"keywords":92,"publishedAt":19},"01M77XD2541C8HXJ6RHYZP4NGJ","youtube-short-maker","Make YouTube Shorts that hold attention with Pexo — vertical 9:16, fast-paced, hook in the first 3 seconds. Describe your topic and Pexo writes the script, generates the shots, picks the models, and assembles a finished Short with music and captions. Use for YouTube Shorts: \"youtube short\", \"make a youtube short\", \"shorts video\", \"short-form video\". NOT for long-form or landscape video.","https:\u002F\u002Fgithub.com\u002FReihantt6\u002Fsns-myagent\u002Fblob\u002Fmain\u002Fagent\u002Fskills\u002Fyoutube-short-maker\u002FSKILL.md",[],[],{"skillNo":94,"name":95,"nameZh":6,"description":96,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":97,"categories":98,"keywords":99,"publishedAt":19},"01M77XCXKVSYK7KTZ3PQBVVGFD","xcode-project-setup","Safely modifies Xcode projects (.pbxproj) to add Swift Packages and link files. Use this skill whenever an iOS project needs dependencies installed (e.g. Firebase, Alamofire).","https:\u002F\u002Fgithub.com\u002FReihantt6\u002Fsns-myagent\u002Fblob\u002Fmain\u002Fagent\u002Fskills\u002Fxcode-project-setup\u002FSKILL.md",[],[],{"skillNo":101,"name":102,"nameZh":6,"description":103,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":104,"categories":105,"keywords":106,"publishedAt":19},"01M77XCR1DV8ZQK2NHWKVY4EWX","x402","Set up Browser Use Cloud payments with x402 — pay per request from a crypto wallet (USDC on Base mainnet), no signup or API key. Two setups it works out up front — \"just use it\" (set up a wallet so you or Claude Code can run cloud browser tasks paid from the wallet — Claude writes and runs throwaway scripts, nothing touches your codebase) or \"build it in\" (install the SDK and write the key + code into your project). Walks through wallet setup, funding, .env, and a ~$1 test run. Use when the user asks about x402, pay-per-use, USDC payments, or wants Browser Use Cloud without an API key. For the free-tier signup (reverse-CAPTCHA → API key), use `browser-use cloud signup` or the `cloud` skill instead.","https:\u002F\u002Fgithub.com\u002FReihantt6\u002Fsns-myagent\u002Fblob\u002Fmain\u002Fagent\u002Fskills\u002Fx402\u002FSKILL.md",[],[],{"skillNo":108,"name":109,"nameZh":6,"description":110,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":111,"categories":112,"keywords":113,"publishedAt":19},"01M77XCMD3R3ZBZQQSW6E0793B","x-twitter-scraper","Build GitHub Copilot workflows with Xquik X API SDKs, REST endpoints, hosted Apify Actor runs, MCP tools, TweetClaw OpenClaw plugin installs, signed webhooks, tweet search, user lookup, follower exports, media actions, and agent automation.","https:\u002F\u002Fgithub.com\u002FReihantt6\u002Fsns-myagent\u002Fblob\u002Fmain\u002Fagent\u002Fskills\u002Fx-twitter-scraper\u002FSKILL.md",[],[]]