AssemblyAI demos Dictation API: custom key terms fix vocabulary, sub-second transcription in 19 languages
AssemblyAI · x · 2026-09-18
AssemblyAI showcases its Dictation API: what the mic hears — "um so like can you send me the uh the Q3 numbers" — comes back as a clean "Can you send me the Q3 numbers?".
Two highlights: a key terms field where you enter vocabulary your users actually say (drug names, SKUs, product names) so those words are spelled correctly instead of phonetically close — the point where most dictation projects die — and sub-second turnaround for typical short clips on Universal-3.5 Pro across 19 languages. Developers are invited to test with their own audio.
More from Multimodal
- QuiverAI launches Arrow 2, its most advanced editable vector graphics model — stuffyokodraws · 2026-09-18
- OpenArt Ad Awards Close Sept 30: $50,000 in Prizes, $15K for Ad of the Year — azed_ai · 2026-09-18
- Prompt recipe makes AI video reveal title letters one by one in cinematic storm scenes — LudovicCreator · 2026-09-18
- Prompt template for letter-by-letter cinematic title reveals in AI video generation — LudovicCreator · 2026-09-18
- Time-travel typography prompt: CHRONOS letters morph through six historical materials — LudovicCreator · 2026-09-18
- Painted keyframes to fluid VFX: open-source IC-LoRA with LTX in ComfyUI — marcosalvi · 2026-09-18