Skip to content

QJX-XXXX/game-visual-forge / skills/forge-text-audio

forge-text-audio

Verified

Generate, redraw, inpaint, continue, validate, and deliver explicitly requested game sound effects and ambience with a locally installed Stable Audio 3 Small-SFX model and Unity AudioClip integration. Use only when the user explicitly asks for non-speech game audio; do not trigger for visual-only work, dialogue, voice-over, songs, or complete musical scores.

references/processing-and-quality.md

View on GitHub
# Processing and quality

Raw provider output and user source files are immutable. Every processed candidate is written to a separate staging directory and binds the request fingerprint, source hash when present, candidate hash, processed WAV hash, waveform hash, and spectrum hash.

The published format is Microsoft WAV, 44,100 Hz, 16-bit signed PCM, mono or stereo according to the confirmed usage and Unity profile. The workflow does not upsample to 48 kHz and does not publish MP3 previews.

The provider applies a no-boost -1.0 dBFS peak ceiling before writing generated raw WAV files. Staging processing may attenuate a still-hot non-looping text-to-audio one-shot to the same ceiling, but it must never raise a quiet signal. Skip silent or already clipped staging input so silence remains reviewable and clipping remains a hard-failure signal. Do not apply upward normalization to any mode.

For one-shots, remove only bounded leading silence and pad or trim the tail to the confirmed duration. For inpainting, convert both source and generated audio to the common PCM format, preserve source samples outside the edit region plus 20 ms guards, and crossfade only inside the guards. For continuation, preserve the converted source prefix before the final 20 ms guard, crossfade into the generated continuation, and enforce the exact target duration. For loops, analyze a 50 ms boundary and apply a 20 ms wrap crossfade without changing sample count.

Hard failures include decode or format mismatch, wrong duration, empty output, clipping, dense full-duration energy without credible one-shot transient-to-tail decay, missing previews, stale hashes, failed protected-region comparison, and missing loop evidence when looping is requested. Speech/music detection is advisory only; the user must still complete the semantic listening check.

Unity profiles are explicit: UI uses mono PCM with preload; one-shots use low-latency PCM or ADPCM without looping; 3D scene sounds use mono; long looping ambience uses stereo Vorbis streaming. The UPM importer writes only under `Assets/GameVisualForgeAudio/<asset-id>/`, preserves existing GUIDs on repeat import, and writes `Reports/unity-import-report.json`. Scene placement is outside the package and occurs only through Unity MCP after an explicit user request.
slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack."Stars: 198kUpdated Claude CodeCodexVerifiedInstall
ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read.Stars: 196kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report.Stars: 196kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt.Stars: 196kUpdated Claude CodeCodexVerifiedInstall