Translation Guide

Explore the translation features of Salad Transcription API. Learn how to translate transcriptions to English, leverage LLM-driven translations between multiple languages, and translate SRT files for captions and subtitles.

Last Updated: October 30, 2024

Introduction

Salad Transcription API not only transcribes audio and video content with high accuracy but also offers exceptional translation capabilities. Our service enables you to:

• Translate from 99 languages to English: Seamlessly transcribe and translate content from virtually any language into English.
• Full translation between multiple languages: Utilize our Large Language Model (LLM) integration to translate between English, French, German, Italian, Portuguese, Hindi, Spanish, and Thai.
• Translating SRT files into different languages.
• Multi-language translation in a single request: Transcribe content and translate it into multiple languages simultaneously.

Translation Features

Translate to English

Automatically translate your transcriptions into English using the "translate": "to_eng" parameter. This is especially useful when working with content in other languages that you want to transcribe and understand in English.

Example:

"input": {
  "url": "https://example.com/path/to/file.mp3",
  "translate": "to_eng"
}

Note: if "translate": "to_eng" is set, the translation will be returned instead of the original transcription. All additional parameters (like timestamps, diarization, etc.) can also be used in conjunction with translation.

LLM-Driven Translation

(Only available in Salad Transcription API)

Leverage our Large Language Model (LLM) integration to translate transcriptions into multiple languages. Specify the target languages using the "llm_translation" parameter.

Supported languages for LLM translation:

• English
• French
• German
• Italian
• Portuguese
• Hindi
• Spanish
• Thai

Example:

"input": {
  "url": "https://example.com/path/to/file.mp3",
  "llm_translation": "german, italian, french"
}

Note: When using “llm_translation”, the original transcription will be returned as normal, along with the translations into the specified languages.

Example Response:

{
  "output": {
    "text": "Hello, and welcome to our conference.",
    "llm_translation": {
      "German": "Hallo und willkommen zu unserer Konferenz.",
      "Italian": "Ciao e benvenuti alla nostra conferenza.",
      "French": "Bonjour et bienvenue à notre conférence."
    }
  }
}

The segments field contains the original transcription in the source language (English in this case). The llm_translation object contains translations into the specified languages.

No need to send a request per language, you can translate your transcription to all supported languages at one operation.

Need output example and a note that original transcription will be return as normal

SRT Translation

(Only available in Salad Transcription API)

To translate your SRT captions into multiple languages, use the "srt_translation" parameter with a comma-separated list of target languages.

Example Request:

"input": {
  "url": "https://example.com/path/to/file.mp3",
  "srt": true,
  "srt_translation": "spanish, portuguese"
}

Note: When using “srt_translation”, the original transcription and SRT content will be returned as normal, along with the translated SRT files.

Example Response:

{
  "output": {
    "text": "Hello, and welcome to our conference.",
    "srt_content": "1\n00:00:00,000 --> 00:00:05,000\nHello, and welcome to our conference.\n",
    "srt_translation": {
      "Spanish": "1\n00:00:00,000 --> 00:00:05,000\nHola, y bienvenidos a nuestra conferencia.\n",
      "Portuguese": "1\n00:00:00,000 --> 00:00:05,000\nOlá, e bem-vindos à nossa conferência.\n"
    }
  }
}

The srt_content field contains the original SRT captions in the source language. The srt_translation object contains the SRT captions translated into the specified languages.

Combining Translation Features

• ** Combining Features**: You can use translation features alongside other parameters like "diarization", "word_level_timestamps", "custom_prompt", "summarize" and others.
• ** Original Transcription**: When using "llm_translation" or "srt_translation", the original transcription and SRT content will be included in the response.
• ** Translation to English**: When using "translate": "to_eng", the transcription is provided directly in English, replacing the original language transcription.

Full Request Example:

curl -X POST salad_url \
-H "Salad-Api-Key: {your-api-key}" \
-H "Content-Type: application/json" \
-d '{
    "url": "https://example.com/path/to/french_audio.mp3",
      "language_code": "en",
      "word_level_timestamps": true,
      "return_as_file": false,
      "diarization": true,
      "sentence_diarization": true,
      "srt": true,
      "sentence_level_timestamps": true,
      "llm_translation": "french, german, portuguese",
      "srt_translation": "spanish"
}'

Example Response:

{
  "id": "b4ed4d73-df5c-43fa-8d2b-b77e57ce5cfc",
  "input": {
    "url": "https://example.com/path/to/french_audio.mp3",
    "language_code": "en",
    "return_as_file": false,
    "diarization": false,
    "sentence_diarization": true,
    "srt": true,
    "sentence_level_timestamps": true,
    "llm_translation": "french, german, portuguese",
    "srt_translation": "spanish"
  },
  "inference_endpoint_name": "transcribe",
  "status": "succeeded",
  "events": [
    {
      "action": "created",
      "time": "2024-10-03T13:58:52.3339361+00:00"
    },
    {
      "action": "started",
      "time": "2024-10-03T13:58:52.417207+00:00"
    },
    {
      "action": "succeeded",
      "time": "2024-10-03T13:59:10.6424658+00:00"
    }
  ],
  "output": {
    "sentence_level_timestamps": [
      {
        "text": "It always seems like time stands still when I'm with you.",
        "timestamp": [29.98, 30.98],
        "start": 29.98,
        "end": 30.98
      }
    ],
    "srt_content": "1\n00:00:29,980 --> 00:00:30,980\nIt always seems like time stands still when I'm with you.\n\n",
    "summary": "Time stands still in love.",
    "llm_translations": {
      "French": "Le temps semble s'arrêter chaque fois que je suis avec toi.",
      "German": "Es scheint immer so, als ob die Zeit stehen geblieben ist, wenn ich bei dir bin.",
      "Portuguese": "Sempre parece que o tempo se detém quando estou com você."
    },
    "srt_translation": {
      "Spanish": "1\n00:00:29,980 --> 00:00:30,980\nSiempre parece que el tiempo se para cuando estoy contigo."
    },
    "text": " It always seems like time stands still when I'm with you.",
    "duration": 0.1,
    "processing_time": 17.778546571731567
  },
  "create_time": "2024-10-03T13:58:52.3339361+00:00",
  "update_time": "2024-10-03T13:59:10.6424658+00:00"
}