Deepgram Flux TTS
#13 v Syntéza řeči (TTS)Deepgram · od 12. August 2026 · 2× · naposledy 17. 10. 2026
34
Momentum
Flux TTS is a streaming-first text-to-speech model from Deepgram for voice agents, served via the /v2/speak endpoint over WebSocket and REST. Unlike conventional TTS, it reads the entire conversation rather than isolated lines and keeps tone consistent across turns. It was released in August 2026; standard pricing began September 13, 2026.
Vývoj momenta
20.07.18.10.
Vlastnosti
| Real-Time Streaming | Yes: streaming over WebSocket /v2/speak, LLM token input, barge-in reports text actually heard |
| Latency | Time-to-first-audio as low as 80 ms (per vendor) |
| Platform | Deepgram cloud API (WebSocket and REST, /v2/speak); also self-hosted, in-VPC or on-prem (dedicated engine requires L4, L40S, A100 or H100 GPUs) |
| Price | $0.0450 per 1,000 characters (Pay As You Go); $0.0405 (Growth). Free through Sept 12, 2026; standard pricing from Sept 13, 2026 |
| Release Date | August 12, 2026 (general availability) |
| Languages | Currently English voices only; more languages on the roadmap |
| Voice Cloning | Not available; on the roadmap |