Two APIs, one developer scorecard, verdict first. Every number traces to the same fact base as the main ranking. Where a vendor publishes nothing, the cell says so rather than guessing.
Deepgram wins on official SDKs and a simple per-character rate; Google wins on SSML and catalogue size. A developer comparison of the two API surfaces and their trade-offs.
ElevenLabs and Cartesia both publish a TTS API, per-plan concurrency, and a commercial licence. This comparison scores the details a developer checks before the first request.
Google and Amazon Polly publish full SSML references; ElevenLabs, Cartesia, OpenAI, and WellSaid use request settings instead. A practical comparison for teams porting a markup pipeline.
Only one vendor in the 2026 run states a latency figure. Why a cross-vendor benchmark misleads, what each API discloses, and how to measure time to first audio yourself.
Google and Polly list $4 per million characters; OpenAI and Deepgram list $15 to $30; Cartesia and ElevenLabs sell credits. This note converts every published rate and shows the formula.