Simplismart

Simplismart hosts open text-to-speech models: Orpheus for English, Veena for Hindi and English, and Qwen 3 TTS on a dedicated deployment. Simplismart tuning is passed in synthesizer.options.

{
  "verb": "say",
  "text": "Hi there, your order ships today.",
  "synthesizer": {
    "vendor": "simplismart",
    "language": "en-US",
    "voice": "tara",
    "options": {
      "temperature": 0.7,
      "top_p": 0.9
    }
  }
}

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.

Credential

Create an API key in your Simplismart account. jambonz sends the key as Authorization: Bearer <api key>. Add it as a Simplismart speech credential in the portal or API, using the vendor id simplismart. Simplismart is used for text-to-speech only.

The credential can also set a default model and an API host; see Models and Endpoint override.

Models

Model	Notes
`canopylabs/orpheus-3b-0.1-ft`	Orpheus 3B, English. Default.
`maya-research/Veena`	Veena, Hindi and English. Default when `synthesizer.language` is Hindi (`hi-*`) and no model is set.
`qwen-tts`	Qwen 3 TTS. Served from a dedicated Simplismart deployment, so set the API host as well.

The model is chosen in this order: options.model_id on the verb, then the model on the credential, then Veena for a Hindi language, then Orpheus. Any model id that contains qwen is sent to Simplismart’s Qwen endpoint; every other model uses the Orpheus/Veena endpoint.

Voices and languages

The portal lists these voices:

Language	Voices
English (US)	`tara` (default), `leah`, `jess`, `mia`, `zoe` (female); `leo`, `dan`, `zac` (male)
Hindi	`kavya` (default), `maitri` (female); `agastya`, `vinaya` (male)

Qwen 3 TTS uses its own voice names, such as Chelsie; enter them as a custom voice.

Orpheus and Veena have no language parameter, so synthesizer.language only picks the default model. For Qwen, jambonz sends a language name derived from synthesizer.language: en English, zh Chinese, ja Japanese, ko Korean, de German, fr French, ru Russian, pt Portuguese, es Spanish and it Italian. Any other language is sent as English. Set options.language to send a different name.

jambonz treats Simplismart audio as 16-bit PCM at 24 kHz and resamples it to the call’s rate.

Options

Every option is optional. An option you don’t set is not sent, so Simplismart’s own default applies.

model_id

string

canopylabs/orpheus-3b-0.1-ft, maya-research/Veena or qwen-tts. Overrides the model on the credential.

Orpheus and Veena

These are sent only to the Orpheus/Veena endpoint and are ignored for Qwen.

temperature

number

Sampling temperature.

top_p

number

Nucleus sampling threshold.

repetition_penalty

number

Penalty for repeated tokens.

max_tokens

integer

Maximum audio tokens generated per request. jambonz sends one request per sentence.

Qwen

These are sent only to the Qwen endpoint and are ignored for Orpheus and Veena.

language

string

Language name, such as English or Chinese. Overrides the name jambonz derives from synthesizer.language.

leading_silence

boolean

Whether Qwen adds silence before the speech.

Endpoint override

To use a dedicated deployment, such as one serving Qwen 3 TTS, set the API host on the credential, or api_uri in its Extra Options or in synthesizer.options. It is a host name such as my-deployment.simplismart.live; a scheme and trailing / are accepted and ignored. Leave it out to use api.simplismart.live.

Streaming

Simplismart has no token-streaming websocket, so it does not support TTS streaming: use it for one-shot say, not for say with stream: true or the agent verb.