SparkDesk 4.0

audio
proprietary
SparkDesk 4.0 is iFlytek's flagship speech recognition model supporting 74 languages. It delivers high transcription acc...
Version: 4.0
Released: 1y 2m 17d ago on 08/15/2024
Pricing:
  • details: see website for pricing
Repository: N/A

Architecture

  • parameters: N/A
  • context_length: N/A
  • inference_type: Transformer-based speech recognition
  • training_data: Multilingual speech corpora with noise augmentation

Capabilities

  • 74 languages
  • high accuracy
  • speech-to-text
  • transcription
  • robust to noise

Benchmarks

  • STT: Competitive with top models (official scores not published)

Safety

  • Standard speech filters (e.g., adult language blocks)
  • No alignment layer (straight transcription)
  • Varies by accent/language coverage

Deployment

  • regions: Global
  • hosting: iFlytek Cloud Platform, APIs
  • integrations: Used in Huawei OceanStor AI storage solution

Tags

speech-recognitionASRmultilingualproprietary

Join our community

Connect with others, share experiences, and stay in the loop.