generate-narration
Generate narration from a script and add it to edit.json audio.narration[]. Read `akari narration engines --json` to choose an engine and voice; use a candidate row only when the user names it. Use for a draft voiceover, final narration, a new voice profile, or replacing an existing voice or engine.
- License 许可
- MIT
- 最近更新
- 2026-10-22
- Stars
- ★ 214
- Forks
- 25
01 前言 Front Matter
- name
- generate-narration
- description
- Generate narration from a script and add it to edit.json audio.narration[]. Read `akari narration engines --json` to choose an engine and voice; use a candidate row only when the user names it. Use for a draft voiceover, final narration, a new voice profile, or replacing an existing voice or engine.
02 资源文件 7
- 资产文件assets/reading-script-v1.md1.4 KB
- 参考文档avatar-voice.md3.4 KB
- 参考文档engines.md14.2 KB
- 参考文档lang/en.md2.9 KB
- 参考文档lang/ja.md5.5 KB
- 参考文档reading-text.md1.8 KB
- 参考文档voice-profile-setup.md8.0 KB
03 SKILL.md 正文
Generate narration audio
Language: Respond in the user's language — conversation, questions, approval prompts, and the reports and records you write (including decision-log rows) follow the language the user writes in.
Read lang/<content-language>.md for language-specific rules (content language: --lang → intake.json "language" → transcript language → ja; see https://github.com/AkariLabs/akari-video/blob/main/docs/contract-2026-10-11-content-language.md). If that file does not exist, read lang/en.md. Pass the work's content language to the TTS engine when its locale interface is available.
Read only the relevant lines of edit.json and captions.json by id, following the edit.json access guide. Write through the Edit operation or the edit-store script API.
FORBIDDEN-level hard rules
These rules take precedence over detailed instructions.
- Clone only the profile owner's own voice. Never use a profile without a consent record.
- Verify reference audio locally before external upload. Run the local Whisper script-matching guard. Skip it only on the owner's explicit instruction and record that exception.
- Generate from prepared reading text. Store both the display script and the exact reading text in edit.json through Edit or the edit-store script API. Follow the content-language pack for language-specific pronunciation.
- Declare the cost, then run paid lanes only after explicit approval. This applies to fal, direct Google AI Gemini, and Fish Audio. Even when a price cannot be estimated, the CLI must stop at exit 2 unless
--yeswas authorized. - Never call an API key directly; use only setup-connections. fal (including Gemini via fal) uses
FAL_KEY, direct Google AI Gemini usesGEMINI_API_KEY, and Fish Audio usesFISH_AUDIO_API_KEY. setup-connections is authoritative for names, sources, and doctor status. Keys live in <!-- F paths.credentials -->~/.akari/credentials.env<!-- /F --> (the legacy <!-- F paths.credentials.legacy -->~/.config/akari-video/credentials.env<!-- /F --> is readable). Do not offer a lane whose doctor status is notok. - Record provenance.provider. For example,
"provenance": {"provider": "voicevox", "credit": "VOICEVOX:ずんだもん"}. Use"provider": "fal"for fal and"provider": "human"for recorded speech. VOICEVOX voices also require character credit. - Send the canonical WAV as a data URI for fal reference audio. Stop before upload if that file exceeds 20 MB.
- Never mix pronunciation or delivery instructions into spoken text. Put only words to be spoken in
--text,--reading-file, or--script-file; do not prefix them with an instruction such as “Read this cheerfully:”. A direct Gemini engine has read an instruction prefix aloud. Create intonation through the script's punctuation, word order, and reading-text conversion; send delivery guidance only through--styleon engines that support a separate API field. See engines.md.
Execution order and leaf files
- Read engines.md to choose a supported adapter and generate the narration. Read the content-language pack before selecting a voice or a draft-to-final workflow.
- Only if an own-voice clone is requested and no profile exists, read voice-profile-setup.md. Present the fixed script, record the voice, check the script match locally, obtain any needed approval, create the clone, save it, and audition it.
- Read reading-text.md to prepare the exact spoken text from the display script, using the content-language pack for pronunciation rules.
- For an avatar, read avatar-voice.md and resolve
laneandreffromvoice/voice.jsonto an engine. - If word timing is needed for the generated audio, for example to align screen motion with words, run
akari media transcribe <音声>through the same silence-aligned L1 path as analyze-footage. Pass the work's language; do not call whisper-cli directly.
Read only the leaf for the current step.
References
- Narration data contract: docs/contract-2026-07-20-edit-json-v1-narration.md
- Audio contract for BGM, SFX, and ducking: docs/contract-2026-07-14-edit-json-v1-audio.md
- Connections, doctor checks, and cost approval (read only; do not edit it): setup-connections
- Existing cloud-transcription implementation: transcribe-cloud.mjs
04 同仓库更多 Skills
verify-project
Verify an AKARI Video project with the L0 / L1 / L2 verification ladder required by an AKARI Video task contract on the current Theia stack. Read when acceptance criteria specify levels such as L0 or L0+L1, to determine exactly what to run and how at each level.(旧名: verify)
verify-edit
Verify edit.json and optional analysis.json / captions.json / media with a deterministic CLI, then finish QA with frame inspection and reporting after PASS. Use immediately after writing or modifying edit.json, before export, and after applying review feedback.(旧名: edit-lint)
setup-remote
Set up remote access to approval reports and previews from another device such as a phone, and send footage to the workspace. Use doctor to assess Tailscale, guide installation and login (human actions), expose the preview server (default 4567) and approval report helper over tailnet-only HTTPS with tailscale serve, connect Taildrop reception to workspace inbox/, and verify access from another device. Use for requests to view reports on a phone, approve remotely, send footage from a phone, or set up remote access. Public internet exposure (funnel) is outside the default scope.
setup-project
Set up an AKARI Video project headlessly. Recursively copy `templates/project-default/`, record the template version, initialize git only when safe, and generate an HTML creation report. No app launch is needed. Use to create a video project or fill in an existing folder as an AKARI Video project.(旧名: create-project)
setup-library
Use for first-time AKARI Video setup or when the current project needs more usable assets. This first-run skill checks ffmpeg / whisper-cli (and can install with approval), reads the bundled catalog with akari assets to propose a starter pack, obtains explicit human approval, then acquires, places, validates, and updates INDEX entries. It also covers the audio library (BGM and sound effects): bulk acquisition of first-party AKARI Sounds from GitHub Releases, a free-source candidate HTML list with drop-folder matching or assisted acquisition for missing families, and keep/drop decisions in the listening gallery. Use for requests such as "I want more BGM" or "add sound effects to the library".(旧名: setup-audio-library)
setup-connections
Set up and inspect AKARI Video connections: AI-service keys, which models are allowed, cost-approval policy, the tool and media-service state, and 3D settings. Use for first-time setup, adding a provider or key later, checking what is usable (「何が使える?」 'what can I use?'), setting a key such as fal (「fal の鍵を設定したい」), 3D settings (「3D の設定を使いたい」), connection checks, and gates before paid generation or external publication. Model selection belongs to the production skills (they read the ledger and preferences); this skill decides allowed models and the cost policy. Works with `.akari/connections.json` and a free, read-only doctor.(旧名: manage-connections)(旧名: connect)