Menu
Use when turning segment text and an authorized voice manifest into recording-needed status or audio artifacts.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
| name | voiceover-adapter |
| description | Use when turning segment text and an authorized voice manifest into recording-needed status or audio artifacts. |
script/voice_segments.jsonvoice/voice_profile_manifest.jsonvoice/enrollment/recording_needed.md, if presentvoice/enrollment/recording_needed.md, when enrollment files are missingvoice/voice_profile_manifest.json statusaudio/voiceover.wav, only when an adapter succeedsaudio/voiceover_manifest.json, only for real generated audiovoice_profile_manifest.json.status as recording_needed when required files are absent.source_text and spoken_text contract so pronunciation fixes do not silently change reviewed captions.动态地 should read de; when an engine may read it as di, rewrite spoken_text to an unambiguous form such as 以动态方式....spoken_text. Examples: replace 更准确地说 with 准确一点说, replace 一个 token 地往后生成 with 逐个 token 往后生成, and replace ambiguous 重新算 with 从头再算 when 重 may be misread.QK^T -> Q 乘 K 转置, sqrt(d_k) -> 根号下 d k, and d_k -> d k.按行归一化 with 对每个当前 token 的那一组分数，分别做归一化 when clarity matters.Attention, softmax, token, ChatGPT, Claude, FlashAttention, GQA, MQA, KV Cache, vLLM, and Multi-Head Attention.Q, K, V, Q/K, KV cache, raw mθ_i, nθ_i, δ_i, or polyphonic phrases such as 长上下文 directly to TTS without a reviewed spoken_text form..wav files.caption_text or visual_text to hide a TTS pronunciation workaround; pronunciation fixes belong in spoken_text.Use when aligning voiceover audio and segment text into subtitle files with readability checks.
Use when validating an episode topic and coordinating deterministic local content-factory workflow setup.
Use when creating, updating, or reviewing global DESIGN.md, global FRAME.md, or an episode-level video_script/FRAME.md for AI paper explainer videos.
Use when composing storyboard, assets, audio, and captions into HyperFrames draft outputs.
Use when shaping hooks, cover headlines, storyboard beats, or visual concepts for paper explainer content that needs a stronger advertising idea without losing factual accuracy.
Use when evaluating episode artifacts and reporting pass, partial, or failed production readiness.