Skip to main content
silk muga 1 is our more expressive model: a hinglish emotion-TTS voice you steer with a global tone and inline events. it is tuned for short, natural, spoken hinglish.

best for

expressive conversational hinglish: reactions, banter, support, storytelling.

model id

muga

​
well supported for

  • language: romanised hinglish — latin-script hindi with english code-mixing.
  • content: short, expressive conversational lines: reactions, banter, support replies, storytelling.
  • emotion: six tones (neutral, happy, excited, sad, angry, whisper) plus inline <laugh>, <chuckle>, and <sigh>.
  • length: 2 to 40 second utterances.
not built for devanagari, other indian languages, heavy regional dialects, or long-form narration beyond ~40s.

​
how it works

  • paragraph tone: prefix your text with one tone tag, e.g. [happy].
  • inline events: drop <laugh>, <chuckle>, or <sigh> where you want them.
  • script: latin script only (romanised hinglish). avoid devanagari.
  • length: best at 2 to 40 second utterances.
see the prompting guide for the full tone list, the tone and event compatibility matrix, and worked examples.

​
tones

tonetagdelivery
neutral[neutral]flat, even, no affect
happy[happy]bright, smiling, mid-energy
excited[excited]loud, fast, pitch-up
sad[sad]slow, breathy, low pitch
angry[angry]tight, clipped, sharp
whisper[whisper]quiet, breathy, no voiced energy

​
example request

curl -X POST https://silk-api.rumik.ai/v1/tts \
  -H "Authorization: Bearer rk_live_•••••••••" \
  -H "Content-Type: application/json" \
  -d '{ "model": "muga", "text": "[happy] Yaar tumne phir wahi joke maara!" }' \
  --output muga.wav

​
parameters

muga is steered by the tone tag in your text. it also accepts the shared sampling parameters. see the API reference for the full schema and a live playground.
fielddefaultnotes
textn/arequired. up to 2000 characters. prefix with a tone.
temperature0.6sampling temperature. 0.7 is reliable for muga.
top_p0.95nucleus sampling.
top_k50top-k sampling.
repetition_penalty1.2penalize repeated tokens.
max_new_tokens2048output length cap.