github vercel/ai @ai-sdk/assemblyai@3.0.85

latest releases: @ai-sdk/workflow-harness@1.0.190, @ai-sdk/vue@4.0.179, @ai-sdk/workflow@2.0.111...
3 hours ago

Patch Changes

  • e6ad386: feat(assemblyai): add the assemblyai-stt-3-7 transcription model and the assemblyai-tts-1 speech model

    • Add assemblyai-stt-3-7, AssemblyAI's latest pre-recorded transcription model, with prompting, keyterms, temperature, and removeAudioTags support.
    • Emit a deprecation warning for universal-3-5-pro, which AssemblyAI has deprecated, and point the universal-2 notice to assemblyai-stt-3-7.
    • Stop listing universal-3-pro, which AssemblyAI has deprecated and no longer includes in its accepted models.
    • Remove the best model and the singular speech_model request parameter it used, which AssemblyAI now rejects. best and slam-1 are sent through speech_models like any other ID, so AssemblyAI's error lists the accepted models.
    • Add assemblyai.speech() for the AssemblyAI text-to-speech API (assemblyai-tts-1), with voice, language, and outputFormat mapping, encoding/container/sampleRate/bitRate provider options, and billed characters reported as speech usage and in providerMetadata.

Don't miss a new ai release

NewReleases is sending notifications on new releases.