Prodigy basic questions


My company is considering to switch from in internal annotation tool we are currently using to prodigy. I have a few questions. Would it be possible to ask them over email or only here?



Hi! We always like answering questions on the forum, because it means that more people can benefit from the answers :slightly_smiling_face: So if it's about general questions about the tool that you couldn't find answers to in the docs or on here, maybe ask the questions here? If it's about the purchasing process or other company-specific details, you can of course also send us an email to

Hi Ines,

Thanks for the answer. I can write them here. I just thought it will be easier for me to share the information with the rest of my team via email.
I have a few questions, hope that is ok.

  1. How do you set up a project in the annotation tool (open the window like you see in the demo)? Do you need to write/run code to do that? Or people without any programming skills can set a project by themselves and start annotating? I couldn't find the documentation about this.
  2. I understand prodigy write the annotated data to a database. Is there an option (in case of text classification for example) to download the annotated data as a csv file in addition to writing it in the database?
  3. Are you planning to support annotations of audio and video files in the near or far future?

Thank you very much,

Hi Ayala,

I hope the following answer your questions:

Normally you would start and stop a task from the command line. The interface is really designed for people with some programming knowledge, because we assume it takes programming knowledge to use the data once it's produced. Even designing a new annotation task takes some knowledge of how the models work and how things fit together -- so we haven't designed the tool to be used independently by non-programmers.

Of course, once the annotation task is started, you can just visit the URL in the browser. So the annotators don't need to know anything about programming.

By default the annotations are saved into a local SQLite database. You can export them into a jsonl format using the prodigy db-out command. Converting that to a CSV format is really easy, but we don't have a built-in for it because the specifics will depend on how you want your CSV data to look.

You can use the HTML view to present audio or video media for annotation, and we do have some users working with audio data. We don't have a lot of custom tooling for this however --- for instance, we don't have a tool that lets you draw bounding boxes for object detection in video. We don't have immediate plans for that.