ELAN (EUDICO Linguistic Annotator) is a free, professional annotation workbench developed at the Max Planck Institute for Psycholinguistics, purpose-built for researchers who need to attach precise, time-stamped labels and transcriptions directly to audio or video recordings. It is the de facto standard in language documentation and corpus linguistics worldwide.
What is ELAN?
ELAN is a linguistic annotation environment that binds every label, transcription, or code you create to an exact time interval inside one or more media files. Rather than writing notes alongside a recording and hoping the timestamps match, you work inside a timeline where each annotation is its time span — drag a boundary and the label moves with the waveform. The underlying file format is XML-based, human-readable, and built to last decades in an archive.
The tool grew out of the field-linguistics community, where a researcher might spend months recording an endangered language and then years transcribing it. That origin story explains everything about ELAN's design priorities: it handles overlapping speakers, simultaneous camera angles, nested analytical tiers from phoneme to discourse, and corpora that run to thousands of annotations without complaint.
What does ELAN do best?
ELAN's defining feature is its multi-tier annotation model. A tier is an independent annotation track — you can stack a transcription tier, a free translation tier, a morpheme gloss tier, a gesture tier, and a tone-marking tier all over the same video, each time-aligned independently. I have opened project files with forty-plus tiers on a single recording and ELAN rendered them without flinching.
- Millisecond-accurate time alignment — annotations snap to waveform transients or exact frame boundaries; playback always lands in the right place.
- Controlled vocabularies — lock a tier to a defined label set to enforce consistency across annotators or across a multi-year corpus.
- Cross-tier search — regex and structured constraint queries that span multiple tiers simultaneously, exportable as concordance tables or frequency lists.
- Deep format interoperability — import and export to Praat TextGrid, Toolbox, FLEx, CHAT, ANVIL, CSV, and subtitle formats. Almost every downstream tool in the field speaks at least one of these.
- Multi-annotator workflow — tier ownership, annotation locking, and file merging for inter-rater reliability studies.
Is ELAN free?
Completely free — no license, no subscription, no feature tier. ELAN is funded as public research infrastructure by the Max Planck Society, which means the core team's incentive is scientific utility rather than conversion rate. It is actively maintained and ships regular updates.
The honest trade-off: ELAN is a Java Swing application, and Retina displays expose that heritage. Text is occasionally small, drag targets are narrow, and new users routinely spend an hour just understanding the tier-parent hierarchy before they annotate a single word. That curve is steep but finite — within a week the workflow becomes muscle memory.
Who should use ELAN?
ELAN's natural home is academic research: field linguists, phoneticians, sign-language specialists, gesture scientists, conversation analysts, and anyone building a searchable, time-aligned spoken corpus. If your work involves transcribing recorded speech at scale — especially in under-resourced or endangered language contexts — ELAN has no credible free rival at this depth.
It surfaces outside academia too: qualitative UX researchers coding think-aloud sessions, documentary teams building detailed shot logs, and oral-history archivists who need annotations that survive format migrations. It is not a general subtitle editor — for that, Aegisub moves faster and feels friendlier.
How does ELAN compare to Praat?
Praat and ELAN are complementary instruments, not competitors. Praat is a phonetician's oscilloscope — it measures formants, pitch tracks, and voice-quality parameters; its TextGrid annotations are tightly coupled to acoustic analysis. ELAN is a transcriptionist's workbench — it handles multi-speaker, multi-modal data at the utterance and turn level and scales gracefully to large corpora.
Many researchers run both in the same pipeline: Praat for fine-grained acoustic coding of individual segments, ELAN for the broader discourse transcription, then import the Praat TextGrids into ELAN to unify the corpus into a single searchable file. ELAN natively reads and writes Praat's format, so the two tools cooperate without friction. Against ANVIL — Max Planck's gesture-centric sibling — ELAN wins on community size, format breadth, and general-purpose flexibility; ANVIL wins on custom display schemes for frame-by-frame gesture coding.
