Google Cloud Speech-to-Text
Speech recognition API on Google Cloud, powered by the Chirp foundation models
About Google Cloud Speech-to-Text
Google Cloud Speech-to-Text is Google Cloud's managed speech recognition API. The V2 API serves the Chirp 3 model alongside older task-specific models through streaming, synchronous, and batch recognition, with automatic language detection across 100+ languages. It bills per second of audio processed and plugs into Google Cloud IAM, regions, and billing.
Pros & Cons
Pros
- Dynamic batch at $0.003 per minute (about $0.18 per hour) is among the cheapest managed batch options
- Chirp 3 covers 100+ languages with automatic language detection
- Fits teams already on Google Cloud: IAM, regional endpoints, Cloud Storage input, and one bill
- Volume tiers drop the standard rate to $0.004 per minute above 2M minutes a month
Cons
- Standard real-time rate of $0.016 per minute (about $0.96 per hour) is several times higher than specialist STT vendors
- Chirp 3 doesn't return word-level timestamps, per Google's model docs
- Diarization is limited to 15 languages and isn't available on streaming
- Most Chirp 3 languages are still in preview, and setup requires a Google Cloud project, recognizers, and IAM
Key Features
Chirp 3 model
Google's latest speech foundation model, covering 100+ languages (about 20 GA, the rest in preview) with automatic language detection via language_codes=['auto'].
Three recognition methods
StreamingRecognize for live audio, Recognize for clips under a minute, and BatchRecognize for long audio from one minute to about an hour.
Dynamic batch
Lower-urgency batch processing billed at a discounted $0.003 per minute.
Speaker diarization
Available on Chirp 3 for Recognize and BatchRecognize in 15 languages; not on streaming.
Model adaptation
Phrase lists of up to 1,000 phrases to boost recognition of domain terms.
Medical models
medical_conversation and medical_dictation models for clinical audio.
Google Cloud integration
US and EU multi-region endpoints, IAM, Cloud Storage input, and consolidated Google Cloud billing and committed-use discounts.
Pricing
- 0-500K min/month: $0.016/min
- 500K-1M min: $0.01/min
- 1M-2M min: $0.008/min
- 2M+ min: $0.004/min
- Includes Chirp 3, latest_long, phone_call, video models
- Billed per second, per channel
- Lower-urgency batch processing
- Same standard models
- First 60 minutes per month free
- $0.024/min without data logging
- medical_conversation and medical_dictation models
- First 60 minutes per month free
Best For
Google Cloud-native apps
Add transcription to products already running on Google Cloud without another vendor contract.
Large batch backlogs
Process archives of recordings cheaply with dynamic batch.
Multilingual transcription
Transcribe audio in 100+ languages with automatic language detection.
Similar Tools
Google Vertex AI
Unified platform for building, deploying, and scaling generative AI and ML models
Groq
Ultra-fast AI inference powered by custom LPU silicon
EKHOS AI
Secure offline AI transcription for legal, medical, and sensitive recordings
Fireflies.ai
The #1 AI notetaker for your meetings
Wispr Flow
AI-powered voice dictation that writes 4x faster than typing
Transgate
Convert audio and video to text in seconds with AI precision
Featured In
Ready to try Google Cloud Speech-to-Text?
Start using Google Cloud Speech-to-Text today and boost your productivity.
Visit Website