Generate and translate AI captions in the Customer Portal

To generate subtitles from the original audio and translate them into the language of your choice, you need to create corresponding tasks for our AI models.

Generate captions from the original video

You can upload any MP4 videos stored on Gcore video hosting or add links to any external MP4 file with HTTP/HTTPS access.

Step 1. Choose a video

1. In the Gcore Customer Portal, navigate to Streaming > AI. The AI tasks page will open.

2. In the Origin URL field, enter the link to your MP4 video. You have two options:

• Paste video origin URL: If your video is stored externally, provide a URL to its location. Ensure that the video is accessible via HTTP or HTTPS protocols.

  To check the example of a correctly formatted URL, use the link under the field. It will autogenerate a sample URL and you can adjust your URL accordingly.

  

• Select from uploaded videos: choose a video hosted on the Gcore platform.

Step 2. Select AI task

In the Task type, choose Transcribe & Translate Captions.

If you want to translate your video into multiple languages, create a separate task for each translation.

Step 3. Select audio language

In the Audio language dropdown, choose the video’s original language. You can keep the default Auto-detect option. The system will identify the language automatically.

If your video contains multiple languages, it’s best not to specify the original language as it might cause incorrect results. Instead, keep the default Auto-detect option so that the system will detect all used languages and prepare subtitles accordingly.

Step 4. Select subtitles language

In the Subtitles language dropdown, select the needed language:

• If you only need transcription, choose the original video’s language.

• If you want to translate subtitles, specify the language in which the subtitles should be generated. Make sure that you explicitly specify the language you want to translate to. Translation into different languages should be done by creating separate tasks.

You can also keep the default Auto-detect option, and the system will create subtitles in the detected language.

Step 5. Generate AI task

After you configure all the requirements, click the Generate task button.

The AI task will appear in the tasks table and will have the Processing status. It might take some time for it to finish, depending on the length of your video. Note that the task might change to Pending a few times, this just means that it’s being queued for another processing stage.

If the status doesn’t change for a while, click Refresh to update the table.

We only charge for the successfully processed translation tasks.

Step 6. Download captions

When the task status changes to Success, you can view and download subtitles. The autogenerated subtitles are in WebVTT format.

To download captions:

1. Find your task in the AI tasks table.

2. Click the task ID to view task details.

3. In the Task details dialog, find the Parsed result section that contains generated and translated subtitles. If you want to download subtitles, click Download VTT file under the Parsed result field.

If a video contains multiple audio tracks (for example, different language tracks or commentary), the service will only recognize and transcribe the first one. We're working to improve the functionality and support processing multiple audio tracks within the video.

AI-generated text may not match the quality of professional manual transcription. If you notice the difference, you can download subtitles and adjust them in an external editor.

AI task statuses

Each AI task is processed in sub-stages. For example, the original language is first determined in a video, and then transcription is done.

In such cases, the video processing status may change from Started to Pending and back. This happens during the allocation of resources for a specific sub-stage. To view a general task processing status, check the progress percentage.

A task can have the following statuses:

• Success: The processing has been completed.
• Started: The processing has started.
• Pending: The task is being processed and is awaiting the allocation of available resources.
• Failure: A task has failed. Create the task again.
• Revoked: The processing was canceled by a user or system.
• Retry: The task execution failed due to internal reasons, the task is queued for re-execution (up to 3 times).

We delete task result data after one month.

Manage AI tasks

All generated AI tasks are available in the Customer Portal on the AI tasks page:

You can check the detailed information about each AI task, sort tasks by creation date, and get the link to the original video file. If you have a large number of AI tasks, you can also use search to find the one that you need.

View task details

To get detailed information about each task, click its ID to open the Task details dialog, which features the following data:

• ID: A unique identifier of an AI task.

• URL: The web address where the processed video file is hosted.

• Task type: Can be either Transcribe & translate captions or Content moderation.

• Status: The outcome of task execution.

• Contributed time: The amount of time spent on task processing.

• Started at, Finished at: The date and time when the task was initiated and when it was completed.

• Parsed result: The output of the AI video processing task along with the WebVTT file that contains transcribed or translated text.

• Request: The API request to the server that initiated the video transcription or translation.

• Response: The respose returned by the server with task details. You might check the "audio_language" and "subtitles_language" parameters to confirm that subtitles have been produced in the correct languages.