Understanding › Translation

Translation

Translate text between languages with LLM-powered translation and real-time streaming output. Send text in any language, name a target language, and read the translation as it's generated over Server-Sent Events.

What you get

Translation streams its response through Server-Sent Events (SSE), so you receive translated text while it is still being generated rather than waiting for the full result.

Included	Typical use case
Real-time translation between languages	Live chat and customer support translation
Streaming output as content is generated	Real-time subtitles and multilingual captions
Source language auto-detection	Cross-language workflows in apps and internal tools
Flexible model and prompt configuration	Tuning translation behavior per product surface

The flow is three steps: send the source text and target language, receive streaming chunks while the translation is generated, then wait for the complete event to get the final translation.

Endpoint

POST /api/v1/translation/stream

Unlike most Munsit endpoints, this one authenticates with either an API key or a JWT token in the Authorization: Bearer header.

Field	Type	Required	Description
`text`	string	Yes	Input text to translate.
`target_language`	string	Yes	Full target language name, such as `Arabic`.
`source_language`	string	No	Source language, or `string` for auto-detection.
`model_name`	string	No	Translation model, default `qwen/qwen3-32b`.
`prompt_name`	string	No	Prompt template, default `quick-translate`.
`prompt_version`	string	No	Prompt version, default `1.2.0`.

Translate with streaming

Keep the connection open and read events until complete arrives.

curl

curl -N -X POST "https://api.munsit.com/api/v1/translation/stream" \ -H "Authorization: Bearer YOUR_MUNSIT_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "text": "Welcome to Munsit", "target_language": "Arabic" }'

Stream events

Four event types arrive over the SSE connection, in order.

Event	What it carries
`session_info`	Request and session metadata.
`start`	Translation start, with the detected source language.
`chunk`	Partial translation content — append as it arrives.
`complete`	Final translation output and metadata.

Keep the connection open. The response is streamed as Server-Sent Events — don't close until you've received a complete event, or you'll miss the final translation output.

Go further

Translation slots into any transcript pipeline.

Transcribe audioProduce Arabic transcripts to translate.POST /speech-to-text Sentiment analysisScore tone and emotion on the same transcript.POST …/sentiment-analysis Keyword extractionPull entities and themes from meeting transcripts.POST …/keyword-extraction AuthenticationAPI keys, JWT tokens, and where each is accepted.guide

Working with an AI assistant? Every page is available as Markdown: add .md to the URL, or send an Accept: text/markdown header. For the whole documentation in one request, point it at llms-full.txt; the page index is llms.txt. Or use Copy Page, top right.