Automatic transcription.
The conversion of spoken audio into searchable text using a speech-to-text (STT) model. Modern systems are neural; the dominant open model is OpenAI Whisper, which Screendog uses.
Why it matters for recordings
A 12-minute recording is opaque — you cannot ctrl-F it. A transcript turns the same recording into a searchable document. Both can be linked from a ticket; the engineer chooses which to consume.
Accuracy realism
Whisper is excellent on clear English audio. Heavy accents, cross-talk, music, and proper-noun product names degrade accuracy. The transcript is a hint, not a court record.
Frequently asked
Are non-English transcripts supported?
Whisper handles many languages; Screendog passes audio through and uses whatever language the model detects.
Try Screendog free.
5 recordings on the free trial. Real Linear, GitHub, Notion, Slack, and Jira filing. No credit card.
Start a workspace — free