简体中文 | English

Text-to-Audio

Synthesizes text into AI voice/voiceover via giggle.pro. Supports multiple voice tones, emotions, and speaking rates.

⚠️ Review Before Installing

Please review the following before installing. This skill will:

1. 1. Write to ~/.openclaw/skills/giggle-generation-speech/logs/ – Task state files for Cron deduplication
2. Register Cron (30s interval) – Async polling when user initiates speech generation; removed when complete
3. Forward raw stdout – Script output (audio links, status) is passed to the user as-is

Requirements: python3, GIGGLE_API_KEY (system environment variable), pip packages: requests

No inline Python: All commands must be executed via the exec tool. Never use heredoc inline code.

No Retry on Error: If script execution encounters an error, do not retry. Report the error to the user directly and stop.

Execution Flow (Phase 1 Submit + Phase 2 Cron + Phase 3 Sync Fallback)

Speech generation typically takes 10–30 seconds. Uses "fast submit + Cron poll + sync fallback" three-phase architecture.

Important: Never pass GIGGLE_API_KEY in exec's env parameter. API Key is read from system environment variable.

Phase 0: Guide User to Select Voice and Emotion (required)

Before submitting, you must guide the user to select voice and emotion. Do not use defaults.

1. 1. Run --list-voices to get available voices:

CODEBLOCK0

1. 2. Display the voice list to the user in a readable format (voice_id, name, style, gender, etc.) and guide them to pick one
2. Ask the user's preferred emotion (e.g. joy, sad, neutral, angry, surprise). Use neutral if no preference
3. Only after the user confirms voice and emotion, proceed to Phase 1 submit

Phase 1: Submit Task (exec completes in ~10 seconds)

First send a message to the user: "Speech generation in progress, usually takes 10–30 seconds. Results will be sent automatically."

CODEBLOCK1

Response example:

CODEBLOCK2

Immediately store task_id in memory (addMemory):

CODEBLOCK3

Phase 2: Register Cron (30 second interval)

Use the cron tool to register the polling job. Strictly follow the parameter format:

CODEBLOCK4

Cron trigger handling (based on exec stdout):

Phase 3: Sync Wait (optimistic path, fallback when Cron hasn't fired)

Execute this step whether or not Cron registration succeeded.

CODEBLOCK5

Handling logic:

• - Returns plain text (speech ready/failed message) → Forward to user as-is, remove Cron
• stdout empty → Cron already pushed, remove Cron, do not send message
• exec timeout → Cron continues polling

View Voice List

When the user wants to see available voices, run:

CODEBLOCK6

The script calls GET /api/v1/project/preset_tones and displays voice_id, name, style, gender, age, language to the user.

Link Return Rule

Audio links returned to the user must be full signed URLs (with Policy, Key-Pair-Id, Signature query params). Correct: https://assets.giggle.pro/...?Policy=...&Key-Pair-Id=...&Signature=.... Wrong: do not return unsigned URLs with only the base path (no query params). The script handles ~ encoding to %7E; keep as-is when forwarding.

New Request vs Query Old Task

When the user initiates a new speech generation request, must run Phase 1 to submit a new task. Do not reuse old task_id from memory.

Only when the user explicitly asks about a previous task's progress should you query the old task_id from memory.

Parameter Reference

Interaction Guide

Before each speech generation, complete this interaction:

1. 1. If the user did not provide text, ask: "Which text would you like to convert to speech?"
2. Must guide user to select voice: Run --list-voices, display list, have user choose. Do not use default voice
3. Must guide user to select emotion: Ask the user's preferred emotion (joy, sad, neutral, angry, surprise, etc.)
4. After user confirms text, voice, and emotion, run Phase 1 submit → Phase 2 register Cron → Phase 3 sync wait