Turn finished webnovel chapters into consistently narrated audiobook audio while preserving the same voice across an ongoing serial.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install webnovel-serial-audio
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install webnovel-serial-audio using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Webnovel Serial Audio?
Webnovel Serial Audio is an Openclaw Skills workflow for transforming completed webnovel chapters into chapter-by-chapter listening audio with one consistent narrator. It prepares clean spoken text, manages pronunciation for recurring names and terms, selects or creates an authorized voice, and submits text-to-speech jobs through the bundled Beatra MCP client.
The skill is designed for serialized production rather than one-off audiobook conversion: it records the current chapter first, supports targeted revisions to unclear blocks, and freezes the selected voice for future chapters. It also includes safeguards for consent, paid clone and synthesis operations, task polling, credit estimates, billing reporting, recovery, and secure automatic updates.
Webnovel Serial Audio Use Cases
- Produce audiobook audio for a newly finished webnovel chapter.
- Build a continuing serial fiction listening library as chapters are released.
- Maintain one narrator voice, tone, speed, and output format across multiple chapters.
- Control pronunciation for recurring character names, locations, invented terms, numbers, and other story vocabulary.
- Use an authorized narrator sample to create a cloned voice after explicit consent and credit confirmation.
- Select a catalog voice when no narrator sample is available.
- Revise only misread or unclear audio blocks instead of regenerating an entire chapter.
- Deliver labeled chapter files in reading order with duration, usage, and net credit charges.
How Webnovel Serial Audio Works
- Build a chapter ledger containing the chapter title, order, block labels, speakable text, and pronunciation table.
- Confirm that the source text is final, readable, and free of chapter headings, author notes, and bracket labels intended only for visual readers.
- Collect pronunciation readings for recurring names, terms, and places. If such names exist without a pronunciation table, stop and request the readings rather than guessing.
- If the user provides a narrator sample, obtain an explicit consent attestation that it is their voice or that the speaker authorized cloning. Upload the clean, single-speaker sample only after consent, then present the live clone admission card, estimate, credit requirements, and short-balance behavior.
- Submit an authorized clone exactly once when the user confirms sufficient credits, or select a catalog voice through
beatra.voices.list. Freeze the resultingvoice_id; never substitute a display name for the voice identifier. - Query text-to-speech model availability with the required capability, confirm voice compatibility and the current-chapter estimate, and synthesize only the current chapter.
- Poll each Beatra task with
beatra.tasks.getuntil it reaches a terminal state, then deliver the labeled chapter audio. - After listening review, edit and resubmit only unclear or mispronounced blocks with new request IDs. Do not regenerate accepted audio or the whole chapter unnecessarily.
- Once the chapter is accepted, reuse the frozen
voice_idfor later chapters, confirm each new paid estimate, and deliver chapters in reading order with actual duration, usage, and billing details. - For failures, recover lost responses with the identical frozen payload and ID, recover unknown task IDs through task listing and retrieval, keep polling on
409, and retry insufficient-balance requests only after the user confirms a top-up.
Webnovel Serial Audio Setup
- Install or open the Webnovel Serial Audio skill package and ensure its bundled
scripts/mcp_client.pyis available. - Complete first-use registration and authorization by following the package references for installation registration and authentication. Do not configure a host Beatra Connector.
- Route every Beatra operation through the bundled MCP client; do not use REST or OpenAPI as a fallback.
- Prepare the final chapter text, chapter number or order, target language, and pronunciation table. Keep each submitted input at or below 50,000 characters and split only at sentence or section boundaries.
- Use MP3 output and speed
1.0by default. Usemodel: "auto"only when every compatible live voice card supports the target language. - For an authorized narrator sample, use a clean single-speaker file of approximately 10–300 seconds and no more than 20 MiB. Upload it through the bundled client:
python3 scripts/mcp_client.py upload ./narrator-sample.wav --mime-type audio/wav
- Before any paid clone or synthesis, confirm the voice, pronunciation table, current-chapter estimate, and available credits. Clone creation and synthesis blocks are paid; planning, casting, and pronunciation preparation are free.
- Manage automatic updates when needed:
python3 scripts/mcp_client.py update --auto off
python3 scripts/mcp_client.py update --auto on
python3 scripts/mcp_client.py update --check
The client checks for updates at most once every 24 hours per installation by default. Updates are verified and fail open, so an update problem leaves the current installation usable and never authorizes a paid-generation retry.
Webnovel Serial Audio Data Schema & Taxonomy
Chapter ledger
| Field | Purpose |
|---|---|
| Chapter title | Human-readable chapter label for delivery |
| Chapter order | Reading and serial sequence |
| Block labels | Stable units for synthesis, review, and targeted revision |
| Speakable text | Clean narration input without headings, author notes, or bracket labels |
| Pronunciation table | Confirmed readings for names, terms, places, and recurring vocabulary |
| Target language | Language used for voice and model compatibility checks |
| Voice ID | Frozen catalog or cloned voice identifier reused across chapters |
| Format and speed | Defaults are mp3 and 1.0 |
Runtime artifacts and metadata
- Uploaded narrator samples are referenced as Beatra artifacts, such as an
artifact_id. - Clone jobs use a display name, explicit
consent_attestedvalue, model selection, and an opaqueclient_request_id. - Synthesis is organized by chapter and block, with a request ID for each paid submission.
- Task IDs are polled until terminal status through
beatra.tasks.get; lost task IDs can be recovered throughbeatra.tasks.list. - Delivery metadata should include chapter order, actual duration, usage, voice consistency review, and
billing.net_charged_credits. - Wallet information is separated into read-only balance data from
beatra.wallet.getand charge history frombeatra.wallet.ledger. - Changed chapter text creates new paid work only for changed blocks and must receive new request IDs; identical frozen payloads and IDs are used only for recovery.
- Audio files should be labeled in reading order so listeners can follow the serial progression.
Webnovel Serial Audio Advanced Features
- Persistent narrator continuity through a frozen
voice_idacross future chapters. - Consent-gated voice cloning with sample validation, live pricing, provisional estimates, and insufficient-balance safeguards.
- Catalog voice discovery using live
beatra.voices.listresults when no sample is supplied. - Capability-aware model selection through
beatra.models.listwithtext_to_speechfiltering. - Block-level revision workflows that minimize cost by regenerating only unclear or misread passages.
- Idempotent recovery using identical frozen payloads and request IDs, plus task discovery for lost task identifiers.
- Explicit handling for
409, terminal task states, cancellation requests, and insufficient-balance errors. - Read-only wallet and ledger checks for current credits, live estimates, and actual charges without inventing top-up operations.
- Safe automatic updates with fixed official discovery and CDN paths, archive and checksum verification, rollback, package-boundary enforcement, and fail-open recovery.
- Reference-driven operation covering workflow ledgers, authentication, registration, task results, billing, MCP diagnostics, updates, and uninstall procedures.
- A focused Openclaw Skills workflow that separates webnovel serial narration from one-shot literary books, course lectures, and personal IP voice-clone episodes.
SKILL.md
Loading
Related Openclaw Skills
txt.by Public Agent Messaging
txt.by enables AI agents to publish, discover, and coordinate through public Markdown messages, threads, profiles, and inboxes.
Helena Marketing Agent
Helena connects OpenClaw to Enrich Labs so an AI marketing agent can research, plan, create, schedule, publish, and analyze cross-channel brand marketing.
Policy Application Advisor
An AI-assisted, human-controlled workflow for discovering, evaluating, preparing, validating, and tracking enterprise policy applications.
Dimensional-Breaking Interactive Playbook | AI-HIVE Original Workflow
A rights-aware AI-HIVE workflow for planning, generating, reviewing, and delivering original animation and game-oriented image and video content.
Wealth Product Talking Clips
Turn an approved wealth product factsheet and authorized stills into separate, compliant talking clips without stitching or inventing product claims.
Used Car Walkaround
Turn one used-car condition sheet into a compliant listing hero still, reusable walkaround script, and one audio-driven inventory video clip.
Featured*
