Term
Speech-to-Text (STT)
Speech-to-Text refers to the automatic conversion of spoken language into text by an AI model. In an AI workflow, STT replaces the keyboard as the input channel — what matters is model size and domain vocabulary.
Speech-to-Text — explained in more detail
STT models listen to raw audio (16-kHz mono is standard) and emit text. Current open-source models such as Whisper (OpenAI), Distil-Whisper or Faster-Whisper run locally on CPU or GPU; commercial APIs like OpenAI, Deepgram or AssemblyAI accept audio and return transcripts. Model sizes range from tiny (~40 MB, very fast, many errors) through small and medium up to large (>1 GB, most accurate). Larger models recognise more languages and technical vocabulary reliably, but require more memory and compute. Weak spots common to all models: proper names, brand names, fast speech, technical jargon, and rare terms.
Example / Practical context
In day-to-day AI work, STT increasingly replaces the keyboard: prompts, code instructions and notes are spoken rather than typed — faster, more natural, and possible in parallel to what is happening on screen. A typical setup: push-to-talk key or foot pedal → audio capture → local Whisper-small → text lands in the editor or directly inside the prompt. Friction arises with domain vocabulary: “Claude” turns into “Kloat”, “boostN” into “Boosten”. Remedies range from initial_prompt hints through a correction layer to upgrading to a larger model.
Distinction from related concepts
A voice assistant (Alexa, Siri) is an end-to-end system combining STT, intent recognition and text-to-speech — STT is only the first stage. Dictation software (e.g. Dragon) is classic STT with a tightly focused profile, often speaker-dependent and trained on the user’s voice. Speech-to-speech models (GPT-4o Realtime, Gemini Live) skip the text intermediate and respond directly to audio — more powerful, but more expensive and harder to embed into custom workflows.
Entdecke mehr
Working with AI: one word, two actions — and an hour of debugging
One word, two meanings: I blocked a Git commit, the AI heard deploy. Why a shared vocabulary decides everything when you work with AI.
GlossarAI Data Analysis
AI data analysis means evaluating data (CSV, Excel and more) via natural language: an AI system writes and runs code behind the scenes to clean, calculate and visualise files — e.g. ChatGPT's Advanced Data Analysis.
LexikonAutomating AI Workflows
No-/low-code automation with Make, Zapier and n8n: building blocks, use cases for agencies and SMBs, pricing models and the costliest pitfalls.