Features and improvements shipping to TOMO
The speech API that powers TOMO's BJJ vocabulary recognition was being retired on May 11. We migrated before the deadline — voice logging continues working exactly as before, no action needed on your end. The 199 most critical BJJ terms are active now, with a full 900+ term expansion planned for a future update.
AssemblyAI retired the speech parameter TOMO relied on for BJJ term recognition on May 11. Without this migration, every recording would have failed silently — you'd speak, TOMO would process, and nothing would come back. We caught it and shipped the fix two days early.
The new speech system has a temporary cap of 200 terms during its beta period, down from the 986 we previously loaded. We hand-picked the 199 terms that generic speech recognition struggles with most — the ones you actually need TOMO to catch correctly.
The 200-term limit is a temporary restriction while the new system is in beta. When AssemblyAI lifts it — or when we move to their next-generation model — TOMO will restore all 986 terms automatically. The full list is preserved and ready to re-activate the moment the limit increases.
What this means for your training: Voice logging works the same way it always has — speak naturally after a session and TOMO transcribes it. The curated 199 terms cover the vocabulary that comes up most in a post-training debrief: positions, submissions, guard systems, and the names coaches and training partners actually use on the mat. The 787 terms temporarily on hold are mostly granular variations (e.g., "armbar from north-south" vs. "armbar") that a good transcription engine handles reasonably even without a boost.