k2

by k2-fsa

FSA/FST framework written from scratch in PyTorch/CUDA.

TL;DR

FSA/FST framework written from scratch in PyTorch/CUDA.

Best for differentiable WFST / lattice training on modern GPUs. Pricing: free.

Category
Open source
License
—
Stars
—
Last push
—
Pricing
free
Platforms
Linux

What it is

k2 is the FSA/FST engine that powers icefall and sherpa. Brings Kaldi-style decoders to PyTorch with autograd. Apache-2.0.

Best for: Differentiable WFST / lattice training on modern GPUs.
Watch out for: Niche audience; requires deep familiarity with FSA decoding.

Install / use

pip install k2

Features

Speaker diarizationNo
Word-level timestampsYes
Streaming / real-timeNo
Languages supported40
HIPAA eligibleNo

k2 vs Whipscribe

Featurek2Whipscribe
CategoryOpen sourceTranscription APIs
Pricingfreefree beta
Speaker diarizationNoYes
Word timestampsYesYes
StreamingNoNo
Languages4099
PlatformsLinuxWeb, API, MCP

Alternatives to k2

OpenAI Whisper
OpenAI

The reference open-source multilingual ASR model from OpenAI.

OSS · MIT ★ 98.1k
whisper.cpp
Georgi Gerganov

C/C++ port of Whisper — runs on anything, from a Raspberry Pi to Apple Silicon.

OSS · MIT ★ 48.8k
faster-whisper
SYSTRAN

4× faster than reference Whisper using CTranslate2 — production sweet spot.

OSS · MIT ★ 22.3k

Whipscribe is a managed faster-whisper + whisperX service. If you want transcripts without running infrastructure, paste a URL or drop a file in the form below — you'll have a transcript in seconds.