Models Overview

Choose the right model for your transcription needs. We offer two model families optimized for different use cases.

Cipher Models

Optimized for batch transcription with excellent accuracy. Best for pre-recorded audio files.

ModelTypeDescriptionCredits/sec
cipher-fastBatchUltra-fast transcription for quick turnaround2
cipher-maxBatchMaximum accuracy for complex audio2

Cipher Capabilities

  • 50+ languages with automatic detection
  • Word-level timestamps
  • Context prompts for domain vocabulary
  • Multiple output formats (JSON, SRT, VTT)

When to Use Cipher

  • Podcast transcription
  • Video subtitle generation
  • Meeting recordings
  • Content that requires word timestamps

Lucid Models

Enterprise-grade models supporting both batch and real-time streaming with advanced features.

ModelTypeDescriptionCredits/sec
lucid-monoBothHigh-accuracy for single language content (English optimized)3
lucid-multiBothMultilingual with code-switching support3
lucid-agentBothOptimized for conversational content3
lucid-liteBothCost-effective for high-volume use3
💡 Tip: For best results with Lucid models, always specify the language parameter. While lucid-multi has good auto-detection, lucid-mono, lucid-agent, and lucid-lite perform significantly better with explicit language specification.

Lucid Capabilities

  • Real-time streaming support
  • Speaker diarization
  • Keywords boost
  • Smart formatting
  • Interim results for live applications

When to Use Lucid

  • lucid-mono: English podcasts, audiobooks, monologues
  • lucid-multi: International calls, multilingual meetings, code-switching audio
  • lucid-agent: Voice assistants, call centers, customer support
  • lucid-lite: High-volume transcription, cost-sensitive applications

Model Comparison

FeatureCipherLucid
Real-time Streaming
Multilingual Support✅ (lucid-multi)
Translation Support✅ (cipher-fast)
Speaker Diarization
Word Timestamps
Context Prompts
Keywords Boost
Smart Formatting
SRT/VTT Output
Interim Results