Generate MP3, WAV, PCM, or OGG_OPUS voiceover audio via Volcengine Doubao TTS 2.0.
npx skills add https://github.com/TeamWiseFlow/wiseflow --skill awk-tts
Please help me install this Agent Skill. Skill: awk-tts Source: https://github.com/TeamWiseFlow/wiseflow/tree/main/skills/awk-tts Command: npx skills add https://github.com/TeamWiseFlow/wiseflow --skill awk-tts
Producing narration audio for videos and content fragments requires a text-to-speech service with proper credentials, voice selection, and quality verification. This Skill wraps the Volcengine Doubao Speech Synthesis 2.0 (seed-tts-2.0) unidirectional streaming API into a single CLI command, handling authentication, voice routing, output placement, and an automatic ASR-based quality self-check.
tts_requirement.md files into MP3/PCM/WAV/OGG_OPUS audio with configurable voice, sample rate, speech rate, loudness, and emotional context.fragments/01-hook/ to read the voiceover copy and voice settings from tts_requirement.md and write speech.mp3 plus a duration metadata JSON into the fragment's artifacts/ directory.Ask the agent to synthesize the voiceover text in your fragment directory into an MP3 using the default Chinese voice, for example: run awk-tts on fragments/01-hook/ to generate the narration audio.
High-intent search queries and answers about installing and using this skill.
Run the awk-tts command with --text or --text-file to synthesize speech through the Volcengine openspeech v3 unidirectional API. Output defaults to an MP3 under a timestamped tmp directory, with a JSON metadata file recording duration and settings.
It accepts either the old console dual-header pair VOLC_TTS_APP_ID plus VOLC_TTS_ACCESS_KEY, or the new console single key VOLC_TTS_APP_KEY. The script prefers the dual-header pair when both are present and never needs the Secret Key.
Official 2.0 voices use the _uranus_bigtts suffix, such as zh_female_shuangkuaisisi_uranus_bigtts (default), plus male, multilingual, and customer-service variants. Cloned voices starting with S_ are routed automatically to the seed-icl-2.0 resource.
Yes, unless --no-asr-check is passed, the audio is transcribed with Volcengine's fast file ASR and compared to the source text using Jaccard similarity at a 0.5 threshold. The result prints PASS or WARN without aborting, and is skipped silently if ASR credentials are missing.
A single synthesis call is limited to 5000 characters, matching the seed-tts-2.0 character-based limit. Longer scripts must be split into multiple calls or fragments before synthesis.
jianshuoConvert WeChat article markdown files into narrated MP4 videos with TTS voiceover and CSS/GSAP animations.
MticoolGenerate subtitles, cut footage, and assemble Final Cut Pro projects as FCPXML and SRT files.
davila7Transcribe and translate multilingual audio using OpenAI's Whisper speech recognition models.
Purple-HorizonsProcess spoken language to text and provide voice responses via OpenClaw.