What the three Cepstral licenses are for

Cepstral gives VICIdial its text-to-speech, but it is commercial software, and getting it working takes more than one license. Integration requires three separate purchases, and each is sold on a different basis. Buy the wrong mix and your TTS either will not run or will choke under concurrent load. Here is what each license is for.

Plan before you buy

Before purchasing, work out two numbers: how many concurrent channels you need, meaning how many lines will be speaking at the same time, and how many dialers the software will be installed on. The text-to-speech software is installed on each dialer that uses it, so your dialer count drives the per-server licenses. Your concurrency drives the per-channel license.

The Voice license

Cepstral offers different speaking voices to suit different needs, and the Voice license is what lets you use one. It is purchased per server. You can buy more than one voice if you need them. One detail trips people up: the voice must be the 8kHz for Linux variant. Telephone audio is 8kHz, so a voice meant for desktop playback will not match the call path your Campaign runs on.

The Channel license

The Channel license sets how many channels Cepstral may use concurrently to generate voice. It is purchased per channel. If ten calls might be hearing freshly rendered TTS (text to speech) at the same instant, you need enough channel capacity to cover that peak. Under-buy and prompts queue or fail when traffic spikes.

The Save-to-file license

The Save-to-file license is an integral part of VICIdial's TTS process, not an optional extra. It is purchased per server. Without it the integration does not complete, even if you have voices and channels in hand. Treat it as mandatory for every dialer running TTS.

The three at a glance

• Voice — per server, choose the 8kHz for Linux voice, buy more than one if needed
• Channel — per concurrent channel, sized to your peak simultaneous TTS lines
• Save-to-file — per server, required for the integration to work at all

Deciding what to buy

The decision flows from your topology: count your dialers for the per-server licenses, count your peak concurrent speaking lines for the channel license, then confirm every dialer has all three before going live.

flowchart TD
  A["Plan TTS deployment"] --> B["Count dialers"]
  A --> C["Count concurrent channels"]
  B --> D["Voice license per server"]
  B --> E["Save-to-file per server"]
  C --> F["Channel license per channel"]
  D --> G["Install on each dialer"]
  E --> G
  F --> G
  G --> H["TTS ready"]

After the licenses

Once Cepstral is licensed and installed, the rest is configuration. To understand what the engine actually does, read what Cepstral is and why VICIdial TTS needs it. To build prompts that speak Lead data cleanly, see controlling TTS pronunciation with SSML. The wider audio prompts and TTS guide covers prompts, voicemail, and the store where rendered audio lands.

Licensing, voice variants, and per-dialer installs add up to real setup work, and the rendered output still has to land in a working Recording format (WAV/MP3) on each box. If you would rather have a managed dialer where this is handled, see our plans and pricing.