Appearance
To build a transcription project on Zooniverse, you use the free Project Builder to upload images as "subjects", design a "workflow" of transcription tasks, write a field guide, and submit for review before public launch. The whole process is no-code; the skill is in task design and clear instructions, not programming. Expect a few days of setup plus a beta-review wait before you go live.
What are subjects, workflows and tasks?
Zooniverse uses three core concepts, and getting them straight saves confusion:
- Subject — a single thing a volunteer sees, usually one image of a page.
- Task — one question or action, e.g. "Transcribe the highlighted line" or "Mark each entry".
- Workflow — the ordered chain of tasks shown for each subject.
Group subjects into subject sets (for example, one set per box or volume) so you can launch in batches and track progress per source.
How do you upload subjects?
For more than a handful of images, use the command-line uploader rather than the web form. Install the Panoptes client and push a manifest:
bash
pip install panoptescli
panoptes configure # enter your Zooniverse credentials
panoptes subject-set upload-subjects 12345 manifest.csvThe manifest is a CSV whose first columns are image filenames and whose remaining columns become subject metadata (catalogue_ref, page_no, parish). That metadata flows straight into your exports, so populate it carefully — it is how you link a transcription back to the original record.
How do you design the transcription workflow?
For handwritten records, the Transcription Task (a marking-plus-text tool) lets volunteers draw a line over each row and type its content. For simpler material, a sequence of short-answer text tasks works well. Keep the workflow under about six tasks; long workflows lower completion. Set retirement so each subject retires after a fixed number of classifications:
text
Retirement rule: retire after 3 classifications (clean print)
retire after 5-7 classifications (difficult hands)Three independent passes is the usual minimum for later consensus.
How do you write the field guide and tutorial?
Zooniverse gives you three help surfaces: a tutorial (pops up on first visit), a field guide (always-available reference panel), and a Talk forum. Use them deliberately:
| Surface | Purpose | Length |
|---|---|---|
| Tutorial | First-run walkthrough | 3-5 steps |
| Field guide | Quick rules and examples | 1 screen per rule |
| Talk | Questions, hard cases, community | Ongoing |
Populate the field guide with real cropped examples of tricky letterforms, abbreviations and dates from your own collection — generic instructions will not survive nineteenth-century secretary hand.
How do you get your data back out?
From the project's Data Exports page, request a classifications export. You receive a CSV where each row is one volunteer's answer to one workflow, with the answers stored as JSON in an annotations column. You then aggregate. The supported route is the Panoptes Aggregation package:
bash
pip install panoptes-aggregation
panoptes_aggregation config exports/workflows.csv 12345 -v 1
panoptes_aggregation extract exports/classifications.csv config.yaml
panoptes_aggregation reduce extractions.csv reducer.yamlThe reduce step produces consensus values per subject. Budget time here: aggregation is where many first-time projects underestimate the effort.
What does the review and launch process involve?
Before public launch your project enters beta review: the Zooniverse team and a panel of experienced volunteers test it and give feedback, often catching ambiguous instructions you cannot see yourself. You can always share a private project URL with your own testers beforehand. Once approved, schedule launch for when you can actively answer Talk posts — early responsiveness sets the tone for the whole community.
Key Takeaways
- Zooniverse's Project Builder is free, no-code, and reviewed before public launch.
- Model your data as subjects, tasks and workflows; group subjects into sets for batched launches.
- Upload at scale with the
panoptescliand a metadata-rich manifest. - Set retirement to 3 or more classifications to enable consensus.
- Fill the field guide with cropped examples from your own collection.
- Plan for aggregation work after export using
panoptes-aggregation.
Frequently Asked Questions
Is Zooniverse free to use for historians?
Yes. The Project Builder is free for not-for-profit research and cultural-heritage projects. You apply for review before public launch, but there is no platform fee.
What is the difference between a subject, a workflow and a task?
A subject is one image (or set), a task is one question you ask about it, and a workflow is the ordered sequence of tasks a volunteer completes per subject.
How does retirement work on Zooniverse?
A subject retires once it reaches a classification count you set (commonly 3-15). Retired subjects stop being shown so the crowd moves on to unfinished material.
Can I get raw data out, or only Zooniverse's aggregates?
You export raw classification CSVs containing every individual answer. Aggregation and consensus are done by you afterwards, typically with the Panoptes Aggregation tools or your own scripts.
Do I need approval before launching publicly?
Yes. Projects go through a beta review by the Zooniverse team and a panel of volunteers before they appear on the public projects list, though you can share a private URL anytime.