AI Speech to Text Dataset and Methodology

Version 2026-07-08-28f9fbe2591d

AI Speech to Text

Scoring data for transcription and speech recognition tools.

Compares transcription quality speed speaker handling export options resilience and cost.

  • 10 providers
  • 9 metrics
  • Updated July 8, 2026
AI Speech to Text score leaderboard
RankProviderOverallBest forHighest metricLowest metricBreakdown
1OpenAI Whisper9.2Open-source high accuracyAccuracy 9.6/10Speed 8.8/10Score details
Score breakdown for OpenAI Whisper
Accuracy9.6/10; 30%; allocated to published score 2.87
Noise Robustness9.4/10; 14%; allocated to published score 1.31
Diarisation9.2/10; 10%; allocated to published score 0.91
Real-time Streaming8.8/10; 10%; allocated to published score 0.88
Speed8.8/10; 10%; allocated to published score 0.87
Speaker Detection9.4/10; 8%; allocated to published score 0.75
Cost Efficiency8.8/10; 8%; allocated to published score 0.70
Punctuation9.2/10; 6%; allocated to published score 0.55
Export Formats9.0/10; 4%; allocated to published score 0.36

Whisper (v4/Turbo) is the gold standard for accuracy and hallucination reduction in 2026.

2Deepgram9.1Real-time & AI AgentsSpeed 9.9/10Diarisation 8.9/10Score details
Score breakdown for Deepgram
Accuracy9.3/10; 30%; allocated to published score 2.73
Noise Robustness9.0/10; 14%; allocated to published score 1.23
Diarisation8.9/10; 10%; allocated to published score 0.87
Real-time Streaming9.9/10; 10%; allocated to published score 0.97
Speed9.9/10; 10%; allocated to published score 0.97
Speaker Detection9.0/10; 8%; allocated to published score 0.70
Cost Efficiency9.2/10; 8%; allocated to published score 0.72
Punctuation9.4/10; 6%; allocated to published score 0.55
Export Formats9.2/10; 4%; allocated to published score 0.36

Deepgram Nova-3 leads with sub-150ms latency, making it the top choice for voice AI agents.

3AssemblyAI9.0Speech IntelligenceSpeaker Detection 9.6/10Cost Efficiency 8.5/10Score details
Score breakdown for AssemblyAI
Accuracy9.2/10; 30%; allocated to published score 2.75
Noise Robustness8.8/10; 14%; allocated to published score 1.23
Diarisation9.1/10; 10%; allocated to published score 0.90
Real-time Streaming8.9/10; 10%; allocated to published score 0.89
Speed8.9/10; 10%; allocated to published score 0.88
Speaker Detection9.6/10; 8%; allocated to published score 0.76
Cost Efficiency8.5/10; 8%; allocated to published score 0.68
Punctuation9.2/10; 6%; allocated to published score 0.55
Export Formats9.0/10; 4%; allocated to published score 0.36

AssemblyAI Universal-2 excels at built-in LLM-on-audio and actionable insight extraction.

4Speechmatics8.8Global AccentsAccuracy 9.4/10Cost Efficiency 8.2/10Score details
Score breakdown for Speechmatics
Accuracy9.4/10; 30%; allocated to published score 2.76
Noise Robustness9.2/10; 14%; allocated to published score 1.26
Diarisation9.3/10; 10%; allocated to published score 0.91
Real-time Streaming8.5/10; 10%; allocated to published score 0.83
Speed8.5/10; 10%; allocated to published score 0.83
Speaker Detection8.8/10; 8%; allocated to published score 0.69
Cost Efficiency8.2/10; 8%; allocated to published score 0.64
Punctuation9.0/10; 6%; allocated to published score 0.53
Export Formats8.8/10; 4%; allocated to published score 0.35

Speechmatics remains the global leader for non-native English and diverse regional dialects.

5Azure AI Speech8.7Microsoft EnterpriseSpeaker Detection 9/10Cost Efficiency 8.4/10Score details
Score breakdown for Azure AI Speech
Accuracy8.9/10; 30%; allocated to published score 2.65
Noise Robustness8.6/10; 14%; allocated to published score 1.20
Diarisation8.8/10; 10%; allocated to published score 0.87
Real-time Streaming8.7/10; 10%; allocated to published score 0.86
Speed8.7/10; 10%; allocated to published score 0.86
Speaker Detection9.0/10; 8%; allocated to published score 0.72
Cost Efficiency8.4/10; 8%; allocated to published score 0.67
Punctuation8.8/10; 6%; allocated to published score 0.52
Export Formats8.8/10; 4%; allocated to published score 0.35

Azure AI Speech is the secure choice for air-gapped enterprise and Microsoft-native stacks.

6Google Gemini Flash STT8.6Multimodal IntegrationSpeaker Detection 9.2/10Noise Robustness 8.4/10Score details
Score breakdown for Google Gemini Flash STT
Accuracy8.8/10; 30%; allocated to published score 2.60
Noise Robustness8.4/10; 14%; allocated to published score 1.16
Diarisation8.5/10; 10%; allocated to published score 0.84
Real-time Streaming9.0/10; 10%; allocated to published score 0.88
Speed9.0/10; 10%; allocated to published score 0.88
Speaker Detection9.2/10; 8%; allocated to published score 0.72
Cost Efficiency8.5/10; 8%; allocated to published score 0.67
Punctuation8.6/10; 6%; allocated to published score 0.51
Export Formats8.6/10; 4%; allocated to published score 0.34

Google Gemini Flash integrates video context for industry-leading context-aware transcription.

7Suno/Bark (STT)8.2Creative AudioAccuracy 8.5/10Punctuation 8/10Score details
Score breakdown for Suno/Bark (STT)
Accuracy8.5/10; 30%; allocated to published score 2.53
Noise Robustness8.2/10; 14%; allocated to published score 1.14
Diarisation8.0/10; 10%; allocated to published score 0.79
Real-time Streaming8.2/10; 10%; allocated to published score 0.81
Speed8.2/10; 10%; allocated to published score 0.81
Speaker Detection8.5/10; 8%; allocated to published score 0.68
Cost Efficiency8.0/10; 8%; allocated to published score 0.64
Punctuation8.0/10; 6%; allocated to published score 0.48
Export Formats8.0/10; 4%; allocated to published score 0.32

Suno's entry into the STT market focuses on emotional prosody and creative audio workflows.

8Otter.ai8.1Meetings & CollaborationSpeaker Detection 8.8/10Cost Efficiency 8.1/10Score details
Score breakdown for Otter.ai
Accuracy8.4/10; 30%; allocated to published score 2.42
Noise Robustness8.2/10; 14%; allocated to published score 1.10
Diarisation8.7/10; 10%; allocated to published score 0.84
Real-time Streaming8.5/10; 10%; allocated to published score 0.82
Speed8.5/10; 10%; allocated to published score 0.82
Speaker Detection8.8/10; 8%; allocated to published score 0.68
Cost Efficiency8.1/10; 8%; allocated to published score 0.62
Punctuation8.4/10; 6%; allocated to published score 0.48
Export Formats8.2/10; 4%; allocated to published score 0.32

Otter.ai remains the user favorite for collaborative meeting notes and automated summaries.

9AWS Transcribe8.0Enterprise AWS PipelinesCost Efficiency 8.5/10Speaker Detection 8/10Score details
Score breakdown for AWS Transcribe
Accuracy8.4/10; 30%; allocated to published score 2.42
Noise Robustness8.2/10; 14%; allocated to published score 1.10
Diarisation8.3/10; 10%; allocated to published score 0.80
Real-time Streaming8.4/10; 10%; allocated to published score 0.81
Speed8.4/10; 10%; allocated to published score 0.81
Speaker Detection8.0/10; 8%; allocated to published score 0.62
Cost Efficiency8.5/10; 8%; allocated to published score 0.65
Punctuation8.2/10; 6%; allocated to published score 0.47
Export Formats8.4/10; 4%; allocated to published score 0.32

AWS Transcribe is the robust workhorse for contact centers and large-scale AWS data pipelines.

10Rev AI7.9Human-verified HybridExport Formats 8.6/10Noise Robustness 7.8/10Score details
Score breakdown for Rev AI
Accuracy8.2/10; 30%; allocated to published score 2.40
Noise Robustness7.8/10; 14%; allocated to published score 1.06
Diarisation8.2/10; 10%; allocated to published score 0.80
Real-time Streaming8.3/10; 10%; allocated to published score 0.81
Speed8.3/10; 10%; allocated to published score 0.81
Speaker Detection7.9/10; 8%; allocated to published score 0.61
Cost Efficiency7.8/10; 8%; allocated to published score 0.61
Punctuation8.0/10; 6%; allocated to published score 0.47
Export Formats8.6/10; 4%; allocated to published score 0.33

Rev AI remains the top choice for those needing a hybrid of AI speed and human-verified precision.

Scoring methodology

Metric weights for methodology 1.0
MetricWeightWhy it mattersTie-break
Accuracy30%The transcript must reliably reflect the spoken words1
Noise Robustness14%Maintains transcription quality outside studio conditions2
Diarisation10%Identifies who said what in multi-speaker recordings3
Real-time Streaming10%Enables live captions and interactive speech applications4
Speed10%Supports practical turnaround for batch and live pipelines5
Speaker Detection8%Detects changes between speakers reliably6
Cost Efficiency8%Enables transcription at scale without runaway cost7
Punctuation6%Improves readability and downstream language processing8
Export Formats4%Supports subtitles timestamps and structured output formats9

Known limitations

Scores reflect a point-in-time editorial assessment and accuracy varies with audio language and environment.

Recent changes

Initial release with 10 providers.

Download this version

Machine-readable access

This release is available in structured JSON and CSV. The versioned URLs are immutable and are the preferred targets for retrieval, citation and reproducible analysis.

Version history