Glossary
An acoustic model is the part of a speech recognition system that maps raw audio (or a mel spectrogram derived from it) to phoneme probabilities — "what speech sounds are here, frame by frame?" In classical ASR (HMM-GMM, HMM-DNN) the acoustic model is a separate trainable component. In modern end-to-end systems like Whisper, the acoustic model is conceptually subsumed inside the encoder.
The classical pipeline
Pre-2017 ASR pipelines had three distinct components: an acoustic model mapping audio frames to phoneme probabilities, a pronunciation lexicon mapping words to phoneme sequences, and a language model picking the most probable word sequence. Each was trained separately. The acoustic model used Gaussian Mixture Models or DNNs over MFCC/filterbank features. Decoding tied them together with weighted finite-state transducers — fast but complicated to train and maintain.
End-to-end replaces it
Modern transformer ASR models (Whisper, wav2vec 2.0, Conformer) combine acoustic, pronunciation, and language modeling into one network trained end-to-end on (audio, text) pairs. There's no explicit phoneme layer — the model learns its own internal representation. This simplifies training enormously and produces better accuracy on real-world audio. The term "acoustic model" is now mostly a historical/research term.
When you'd still encounter the term
Niche use cases: low-resource languages where you bootstrap from phoneme-aligned corpora; pronunciation dictionaries for forced alignment; research on disentangling acoustic vs linguistic biases. For commercial offline transcription (what we do), the term is mostly informational — you get a transcript out, you don't tune an acoustic model.
In practice
Pre-Whisper (2018), a typical Kaldi pipeline ran a chain-trained TDNN acoustic model producing senone (sub-phoneme) probabilities at 30 ms intervals, then composed those with a 4-gram language model in a WFST decoder. Whisper does all of that implicitly inside its encoder-decoder transformer — no separate acoustic model exposed.
Related terms
Further reading
Try modern ASR
Get started — freeLast updated: September 20, 2026
Frequently Asked Questions
What does Acoustic model mean in transcription?
The component of an ASR system that maps raw audio waveforms to phoneme probabilities.
Why does Acoustic model matter when choosing a transcription workflow?
Acoustic model affects how you understand transcript quality, timing, compatibility, or the technology behind speech-to-text results.
Where can I apply Acoustic model on Transcript.you?
Use the related workflow at /ai-transcription when you want to see how this glossary concept connects to an actual transcription task.
Browse the full transcription glossary or read the complete guide to AI transcription.