Resemble

Resemble tuning is passed in synthesizer.options.

1 {
2   "verb": "say",
3   "text": "Thanks for calling, how can I help you today?",
4   "synthesizer": {
5     "vendor": "resemble",
6     "voice": "your-voice-uuid",
7     "options": {
8       "project_uuid": "your-project-uuid"
9     }
10   }
11 }

You can also set these in the portal under Extra Options on the speech credential, in which case they apply to every request that uses that credential. Settings supplied on the verb take precedence.

The voice value is a Resemble voice uuid. Resemble is text-to-speech only, so the Use for STT option is not offered.

Options

project_uuid

string

Uuid of the Resemble project the generated clip is stored in. Optional — omit it unless you want synthesised clips filed under a particular project in your Resemble account.

use_hd

boolean

Higher-definition synthesis, at a small latency cost. Defaults to false.

use_hd only takes effect on the non-streaming synthesis path — that is, when TTS streaming is disabled or when jambonz is pre-rendering audio for the cache. Resemble’s streaming websocket request does not carry it, so on the default streaming path setting use_hd has no effect.

On-prem container

Resemble credentials can point at a self-hosted Resemble container instead of Resemble’s cloud. In the portal, check Use on-prem Resemble container and supply the TTS Container URI, plus Use TLS if the container terminates TLS. Leave the checkbox clear to use f.cluster.resemble.ai.

Streaming

Resemble supports TTS streaming. Two behaviours are worth knowing about when reasoning about latency:

Resemble has no cancel message, so a barge-in is handled by dropping and re-establishing the connection.
The vendor closes the connection after each completed utterance, so each turn uses a fresh connection.

Both are handled for you by jambonz; they cost a connection setup per turn.