Audio & Video

Label audio and video files in-house for tasks such as speaker diarization and classification. Use Prodigy’s fully scriptable back-end to build powerful workflows by putting your model in the loop.

Fast audio labelling

Highlight slices of an audio, and mark them with whatever labels you need. You can zoom in and out to make your annotations more precise, and easily start, stop and replay the input to make sure you get it right. Keyboard shortcuts and scriptable automation make sure you'll fly through the task as quickly as possible. Perfect for tasks like speaker diarization, topic segmentation and disfluency elimination.

Read more

Try it live and select regions!

This live demo requires JavaScript to be enabled.

Try it live and select regions!

This live demo requires JavaScript to be enabled.

Flexible video annotation

You can use the same interface to highlight slices of a video file too! Even if the video does not contain audio, you can re-use the same interface to select parts of the stream.

Once spans of the video are selected you are free to post-process them as you see fit. Want to re-use the annotated video segments to create images for computer vision? Prodigy returns the annotations via db-out in a convenient jsonl format that's easy to integrate into other systems by just writing a Python script.

Read more

Classify, summarize or transcribe

Quickly sort audio or video inputs into custom categories, marking them as relevant or irrelevant, or grouping them for further analysis. Add one or more text boxes to transcribe audio or summarize scenes, with simple server-side validation to make sure your data meets whatever consistency and quality standards you require.

Read more

Try it live and select the category!

This live demo requires JavaScript to be enabled.

Try it live and transcribe the audio!

This live demo requires JavaScript to be enabled.
recipe.py@prodigy.recipe("custom-audio-recipe")
def custom_audio_recipe(dataset, data_dir):
    stream = Audio(data_dir)
    model = load_your_model()
    return {
        "dataset": dataset,
        "stream": model(stream),
        "update": model.update,
        "view_id": "audio_manual"
    }

Plug in your own models

Custom recipes let you integrate machine learning models using any framework of your choice, load in data from different sources, implement your own storage solution or add other hooks and features. No matter how complex your pipeline is – if you can call it from a Python function, you can use it in Prodigy.

View examples
View the documentation