Diffgram
Open source training-data platform unifying annotation, dataset catalog and workflow automation.
About
Diffgram is an open source training-data platform with three parts: an annotation studio, a dataset catalog for querying and streaming data, and a workflow layer for automation. It supports image, video, 3D, text, audio and geospatial annotation, with ML integrations for pre-labeling and automation, and it can be self-hosted or run on Diffgram's cloud. The project has around 1,900 GitHub stars and was founded in 2018.
The pitch is consolidation: one tool replacing a patchwork of isolated labelers, with unlimited annotations and automations on paid tiers. The core is source-available under a commercial open-source license, and an enterprise edition adds support and managed options. Caveat: community activity is modest for the category and public pricing is not published, so teams should validate support terms and release cadence before committing.
What it can do
- Annotation for image, video, 3D, text, audio and geospatial
- Dataset catalog with query and streaming
- Workflow automation and ML pre-labeling integrations
- Unlimited annotations and automations on paid tiers
- Cloud or on-premise deployment
Setup
Self-host via the documented install, or try the online playground and book a discovery call for enterprise.
Prerequisites
Docker-capable host for self-hosting; cloud storage or local volumes for datasets.
Use cases
Works with
Trust score — 6/10
Scored 6/10 (metrics tier, metrics-v1). Evidence: Strengths: 1911 GitHub stars (verified via API, pushed 2026-06-22), founded 2018, self-hostable, multi-modal annotation; annotation + catalog + workflow in one platform; cloud and on-premise; Docked for: (1) modest community activity for the category; (2) no public pricing; (3) enterprise terms unverified. Manual review: caveat on release cadence and support terms noted in entry. License: commercial open-source license (source-available; enterprise edition separate).
★ 1,911 GitHub stars
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 6.0/10
- Maintenance · recency, releases, issue load
- 6.0/10
- Openness · license + docs
- 8.0/10
- Manual review · sanity read + red-flag check
- 6.0/10
- · Strengths: 1911 GitHub stars (verified via API, pushed 2026-06-22), founded 2018, self-hostable, multi-modal annotation
- · annotation + catalog + workflow in one platform; cloud and on-premise
- · Docked for: (1) modest community activity for the category; (2) no public pricing; (3) enterprise terms unverified
- · manual review: caveat on release cadence and support terms noted in entry
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.