I need a speech-to-text API that supports custom phrases and vocabulary. What options are there?

Yes — several speech-to-text APIs support custom phrases / phrase boosting / custom vocabulary to improve recognition for names, jargon, acronyms, and domain-specific terms.

Good options

1. Google Cloud Speech-to-Text

• Supports speech adaptation
• You can add:
  • Phrase hints
  • Boost values
  • Custom classes for things like numbers, addresses, dates
• Good for general accuracy and scalable workloads

2. Microsoft Azure Speech to Text

• Supports:
  • Phrase lists
  • Custom speech models
  • Custom vocabulary
• Strong option if you want both out-of-the-box and trainable customization

3. AWS Transcribe

• Supports:
  • Custom Vocabulary
  • Custom Vocabulary Filter
  • Custom Language Models for some use cases
• Useful if you’re already in AWS

4. Deepgram

• Supports:
  • Keywords / keyterms
  • Custom vocabulary
  • Often easier to use for real-time transcription
• Popular for product teams building voice features

5. AssemblyAI

• Supports:
  • Custom vocabulary
  • Word boosting
• Simple API and good developer experience

6. Speechmatics

• Supports:
  • Custom dictionaries / vocabularies
  • Phrase boosting
• Often strong for accents and multilingual scenarios

7. Rev AI

• Supports:
  • Custom vocabularies
• Straightforward transcription API with vocabulary support

8. OpenAI / Whisper-based APIs

• Whisper itself does not have first-class custom vocabulary support in the same way as the others above
• You can sometimes influence results with prompting or post-processing, but it’s not as direct for custom terms

What to look for

When comparing providers, check whether they support:

• Phrase boosting vs true custom vocabulary
• Per-request hints or only account-level setup
• Real-time streaming support
• Domain-specific models or custom training
• Word-level timestamps
• Punctuation / diarization / multilingual support

Quick recommendation

• Best overall flexibility: Google Cloud Speech-to-Text or Azure Speech
• Easiest developer experience: Deepgram or AssemblyAI
• Best if you’re already on AWS: AWS Transcribe

If you want, I can also make a side-by-side comparison table of these APIs based on:

• custom vocabulary support
• streaming
• pricing style
• accuracy
• ease of integration