regn.ioβs Supreme 1 is a speech-to-text model and API service focused on English telephone audio. Its goal is not to cover every audio scenario, but to process rough telephony audio at low cost, including GSM, G.711, PCMU/PCMA, low-bitrate Opus, and call-center WAV files. It supports batch HTTPS uploads and real-time WebSocket streaming transcription, with outputs available as plain text, rich text, JSON, CSV, segments, and timings.
The model is the proprietary Supreme 1. The page states a size of 134M parameters and supports GPU FP32/FP16 as well as CPU INT8 execution. The advertised speed is 150-500x, while the real-time interface is listed at 65x realtime. The API design is fairly engineering-oriented: cold regions may return 202 queued, job_id, poll_url, and route_status, and clients need to poll for task status. Its explicit limitation is that it is βEnglish telephony first,β so it is not suitable for clean studio recordings, media audio, or multilingual speech. Chinese support is not disclosed and should be considered unclear or unsupported.
Pricing is the biggest highlight: Supreme 1 costs $0.000099/minute and provides 2 hours of free WAV testing per IP. The page also compares it with whisper-1 at $0.006/minute, emphasizing its cost advantage for large-scale telephone recordings. The hosted version has Standard, Optimized, and Accelerated tiers; private cloud or on-premises deployment requires a custom quote.
Its strengths are extremely low pricing, a clear focus on telephone audio, support for both batch processing and real-time streams, and rich output formats. The documentation also explains routing, cold starts, scaling, and polling mechanisms in detail. The downsides are its narrow application scope, possible waiting time from cold starts, and the lack of sufficient disclosure in the main content regarding privacy policy, payment methods, support channels, and company location. It is suitable for call centers, customer-service QA, voicemail archiving, and teams doing real-time transcription of English phone calls; it is not a good fit for general use cases such as Chinese speech, meeting recordings, or multilingual media subtitles.
Access status from mainland China is not stated, and network availability and payment methods are also unknown. If your project requires Chinese speech recognition, domestic compliance, or low-latency access, you should also evaluate speech recognition services from Chinese cloud providers. If you only process English telephone audio, regn.io can be considered as a low-cost API alternative, but you should first use the free WAV test to verify recognition quality and connectivity.
β This review is compiled from public sources and does not constitute a purchase recommendation. Verify all facts on the vendor's official site. Verify on regn.io official site.
regn.io is an Unknown AI Apps provider. TG4G tracks its product information, an overall rating of 7.0/10, and a China-accessibility score of Workable. Click "Visit Official Site" to reach regn.io directly.