Cartesia provides text-to-speech (Sonic), speech-to-text (Ink) and voice agent APIs for developers, with voice cloning on paid plans. Free plan has 20K credits a month; Pro is $5 a month.
Deepgram offers speech-to-text, text-to-speech and Voice Agent APIs for developers, in the cloud or self-hosted. It advertises $200 in free credit, then pay as you go from about $0.0043 per minute for pre-recorded audio.
Deepgram is a voice AI company that sells speech technology to developers and enterprises through APIs. The home page lists four product areas: speech-to-text for real-time transcription, text-to-speech with a model called Flux TTS, a Voice Agent API that combines transcription, a language model and speech in one connection, and an Audio Intelligence API that extracts information from conversations.

Deployment can be in the cloud or self-hosted, and processing can be real time or batch. The footer links to documentation, a changelog, an API playground, a community and self-hosting guides. The home page does not list supported languages or compliance certifications, and a link to its security page could not be fetched, so check those directly if you have regulatory requirements.
Pricing starts with a $200 free credit. The Pay As You Go plan has no minimum spend and the credit does not expire. Pre-recorded speech-to-text is billed per minute: Nova-3 monolingual is $0.0043 per minute, Nova-3 multilingual $0.0052 and Whisper Large $0.0048. Streaming rates are shown at promotional levels that the page labels as limited time, for example Nova-3 monolingual at $0.0048 compared with a regular $0.0077. Text-to-speech starts at $0.015 per 1,000 characters for Aura-1. Voice Agent calls run from $0.05 to $0.163 per minute depending on configuration. Growth plans of prepaid annual credit start at $4,000 per year, and Enterprise is custom.
It is a tool for building products, not a transcription app, so expect to use an engineer or a no-code connector.
It advertises a $200 free credit. The page does not say whether a card is required for the credit, though Pay As You Go states none is needed.
Pre-recorded Nova-3 monolingual is $0.0043 per minute. Streaming prices shown are limited-time promotional rates.
Yes. The site lists a self-hosted deployment option alongside its cloud service.
Website: https://deepgram.com/
Company: Deepgram
Developer API for speech-to-text, text-to-speech and voice agents, with a self-hosted option. $200 free credit; pre-recorded from $0.0043/min. Needs engineering work.
| Pricing model | Usage-based |
|---|---|
| Free tier | Yes · $200 free credit, then pay as you go with no expiry |
| Starting price | Not verified |
| Data residency | Not verified |
| Compliance (as stated by vendor) | Not verified |
| Integrations | Not verified |
| Public API | Yes |
| Self-hostable | Yes |
| Official site | Online (auto-checked 2026-10-12) |
| Can a general AI assistant replace it? | General LLMs do not replace a speech API for bulk or live transcription; pick Deepgram when you must process many hours of audio. |
| Last verified | 2026-10-10 by editor |
Cartesia provides text-to-speech (Sonic), speech-to-text (Ink) and voice agent APIs for developers, with voice cloning on paid plans. Free plan has 20K credits a month; Pro is $5 a month.
The OpenAI API is OpenAI's developer platform for text, image and audio models, billed per use and separate from ChatGPT. Speech-to-text is listed from $0.0045 per minute; no free credit is mentioned on the pricing page.
Amazon Polly is the AWS text-to-speech service with Standard, Neural, Long-Form and Generative voices, billed per character. The free tier covers 5 million Standard characters a month; Neural voices cost $16 per million characters.
Azure Speech is Microsoft's cloud speech service for text to speech, speech to text, translation and voice agents, now sold as part of Microsoft Foundry. The free F0 tier includes 0.5 million neural text-to-speech characters a month.
WellSaid is an AI text-to-speech platform for voiceovers built with professional voice actors, with an API and 50+ languages and accents. The free trial gives 3 minutes a month; Starter is $19 a month.
Typecast is an AI voice generator with 700+ voices, voice cloning, a web studio, iOS and Android apps and an API. The Free plan gives about 5 minutes of lifetime downloads; Basic is $5 a month.
Need the outcome rather than the software?