More AI tools →74 free dev tools →
- Wire-1: 50 ms p95 to final transcript
- 3.0% word error rate on accented English
- 120 free hours, $0.25 per audio hour thereafter
- Works with OpenAI SDK and any /v1/listen‑compatible client
- Integrations for Pipecat, LiveKit, Vapi, Vercel AI SDK
Real-time English speech-to-text API with sub-50 ms latency and pay-as-you-go pricing.
- for
- Developers building voice agents, contact-center bots, live captions and meeting-note tools.
- pricing
- freemium · free trial
Key features
- Sub-50 ms latency — 34 ms median (p95 50 ms) to final transcript on independent benchmark.
- Streaming API — WebSocket streaming returns interim words and a final punctuated transcript with end-of-turn signal.
- File transcription — POST WAV, MP3, FLAC, Ogg or WebM up to 60 min and get a transcript back.
- OpenAI-compatible endpoint — Works with official OpenAI SDKs via /v1/audio/transcriptions after changing base URL.
- /v1/listen compatibility — Drop-in replacement for existing /v1/listen clients by changing the endpoint URL.
- Security & privacy — Audio processed in memory, never stored; connections encrypted via HTTPS/WSS; request IDs for support.
Use cases
- Voice agents that need a reply within milliseconds after a user stops speaking.
- Contact-center bots streaming phone audio (μ-law, A-law, PCM) in real time.
- Meeting-note services uploading recorded files for fast transcription.
- Live broadcasts such as press conferences transcribed as they happen.
- Live captioning where interim words appear while the speaker talks.
Saidwire pricing
- Free$0120 free audio hours · No card needed · Up to 4 concurrent streams
- Pay as you go$0.25 per audio hour/secondReal-time streaming English · Batch files up to 60 min · OpenAI-compatible · Audio never stored
- Starter$25one payment100 audio hours usable anytime
- Growth$99 a monthmonthly500 audio hours per month · Up to 20 concurrent streams · Priority email support
- Pilot$500one payment30-day pilot · Dedicated key · Setup help · Results summary
Saidwire vs alternatives
Otter.ai | Vonage | Twilio | |||
|---|---|---|---|---|---|
| Best for | Ultra-low-latency English STT | General meeting transcription | Self-hosted STT | Full communications platform | Comprehensive telephony APIs |
| Pricing | Freemium | Subscription | Free | Subscription | Subscription |
| DevHunt upvotes | 1 | 0 | 0 | 0 | 0 |
| Launched | Nov 2026 | — | — | — | — |
- Saidwire vs Otter.ai: Otter.ai provides multi-language transcription and meeting-note UI, but higher latency and different pricing.
- Saidwire vs OpenAI Whisper: OpenAI Whisper is an open-source model you run yourself, whereas Saidwire is a hosted low-latency API.
- Saidwire vs Vonage: Vonage offers broader voice, SMS and video APIs; its STT is not focused on sub-50 ms latency.
- Saidwire vs Twilio: Twilio provides voice call handling and STT as part of a larger suite, but not the same ultra-low latency guarantee.
Saidwire FAQ
What do I get on the free trial?+
120 free audio hours, no credit-card required, and up to 4 concurrent streams.
Is the API always on?+
Yes, the API has no warm-up window and is available 24/7.
Can I use the OpenAI SDK?+
Yes, for whole-file transcription; set the base URL to https://api.saidwire.com/v1 and call audio.transcriptions.create.
Which languages and audio formats are supported?+
English only; streaming accepts linear PCM (8-48 kHz) and μ-law/A-law phone audio, files can be WAV, MP3, FLAC, Ogg or WebM up to 60 min.
How fast is the service?+
34 ms median (p95 50 ms) to final transcript on Coval’s benchmark, ranked 1st of 29 for time-to-final segment.
Do you store my audio?+
No. Audio is processed in memory and never stored; transcripts are not logged.
Summarized by DevHunt from saidwire.com · Nov 11, 2026. Details may change; check the official site.
Otter.ai
Vonage
Twilio




