メニュー
Generate audible speech from text using a TTS API and play it back. Invokes ./scripts/speak.sh to convert text to a WAV file, manages a rolling buffer of 5 recent audio files, and plays the result.
デフォルトでは、最初にソースを確認する Prompt が選択されています。直接コマンドに切り替えるか、ローカルコピーをダウンロードすることもできます。
インストールを決める前に、SKILL.md と SkillsMP に表示されている付属ファイルをお読みください。
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
直接コマンドでは確認用 Prompt が省略されます。実行前にソースを確認してください。
npx skills add https://github.com/SorraTheOrc/SorraAgents --skill speakコマンドは1行のまま表示されます。コピー前に横へスクロールして全体を確認してください。
ローカルで確認しますか？SkillsMP が現在取得できるファイルをダウンロードできます。
EXECUTE immediately when invoked via /skill:audit. Do NOT ask permission or offer alternatives. Provide concise project / work item status and run Worklog helpers to augment results. Trigger on user queries such as: 'What is the current status?', 'Status of the project?', 'What is the status of <work-item-id>?', 'status', 'status <work-item-id>', 'audit', 'audit <work-item-id>'
Canonical dev-to-main release workflow. Provides the release process (dev → main merge) with automated gating for unmerged branches, audit readiness, critical items, worklog refs, and producer review. Trigger with: /skill:ship release
Discover related work for a Worklog work item and generate a concise, auditable "Related work" report that can be appended to the work item description.
SKILL.md を表示中
| name | speak |
| description | Generate audible speech from text using a TTS API and play it back. Invokes ./scripts/speak.sh to convert text to a WAV file, manages a rolling buffer of 5 recent audio files, and plays the result. |
Generate audible speech from text input using a Text-to-Speech (TTS) API,
save the resulting WAV file to <repo-root>/.pi/speak/ with a rolling buffer
of 5 files, and attempt playback.
Invoke the underlying bash script directly:
./scripts/speak.sh "Text to convert to speech"
Or set the TTS_API_URL environment variable to use a different endpoint:
TTS_API_URL="http://localhost:8000/v1/audio/speech" ./scripts/speak.sh "Hello"
| Argument | Description |
|---|---|
text | Text string to be spoken (enclose in quotes for multi-word phrases) |
--stream | Write raw audio to stdout (pipe over SSH to a remote player) |
--help | Show usage instructions and exit |
| Variable | Default | Description |
|---|---|---|
TTS_API_URL | http://100.79.231.101:8000/v1/audio/speech | Override the TTS API endpoint |
SPEAK_DIR | <repo-root>/.pi/speak/ | Override the output directory |
Generated WAV files are stored in .pi/speak/ with the following naming:
speech.wav (most recent)speech.1.wavspeech.2.wavspeech.3.wavspeech.4.wav (oldest retained)When a 6th file is generated, speech.4.wav is removed first, ensuring at
most 5 files are retained at any time.
The script attempts playback in the following order:
termux-media-player (Termux/Android) -- preferred when running under Termuxpw-play (PipeWire) -- preferred on modern Linux desktopsaplay (ALSA) -- fallback for older Linux or minimal environmentspowershell.exe (WSL) -- fallback when running under WSLcmd.exe /c start (WSL) -- final fallback under WSLPlayback failure is non-fatal: the WAV file is still generated and saved even if no audio player is available or playback fails.
The script calls an OpenAI-compatible /v1/audio/speech endpoint with a JSON
payload:
{
"model": "tts-1",
"input": "Text to speak",
"voice": "alloy"
}
A 60-second timeout is applied to the API call.
| Code | Meaning |
|---|---|
| 0 | Success (WAV generated; playback may have failed) |
| 1 | Error (missing argument, API failure, curl error) |
| 0 | Stream mode: audio written to stdout (pipe to aplay etc.) |
# Basic usage
./scripts/speak.sh "Hello, world!"
# Multi-word phrase
./scripts/speak.sh 'The TTS system is now working.'
# Stream raw audio to stdout (pipe over SSH to local player)
./scripts/speak.sh --stream "Hello" | aplay
ssh user@host "cd ~/.pi/agent/skills/speak && ./scripts/speak.sh --stream 'hi'" | aplay
# Custom API endpoint
TTS_API_URL="http://localhost:8000/v1/audio/speech" ./scripts/speak.sh "Test"
# Custom output directory
SPEAK_DIR="/tmp/my-speech" ./scripts/speak.sh "Custom output"
./scripts/speak.sh -- the underlying implementation script