eachlabs-voice-audio

Audio transcription with speaker diarization converts spoken audio into text using models like Whisper or ElevenLabs Scribe, adding word and segment timestamps to identify individual speakers. This process generates speaker-aware text output for meeting transcription.

Text-to-speech generation uses models like ElevenLabs, Kling, PlayAI, Kokoro, and Stable Audio to synthesize natural speech from text scripts. You post prediction requests with schema-matching inputs and poll until completion to receive the generated audio output.

Voice conversion and cloning transform spoken audio using models like RVC and voice changer utilities. This allows you to modify, dub, or clone voices for media production, outputting transformed audio files through authenticated prediction API calls.

Yes, you need an EACHLABS_API_KEY header to authenticate model discovery and prediction calls. The skill validates model existence via the model endpoint, posts prediction requests, and polls status until completion to return the final audio or transcription outputs.

Speech-to-text transcription supports multilingual audio processing through models like Whisper, faster Whisper variants, and ElevenLabs Scribe. These models convert spoken audio into accurate transcripts, applying to scenarios like multilingual narration creation.

Polling for audio prediction completion requires waiting for the model endpoint to return a finished status before retrieving outputs. This asynchronous approach means you must continuously check prediction status until the text-to-speech or speech-to-text workflow is fully complete.