Skip to content

Fast & private Try it Get the apps Pricing Library Setup guide Get it for Mac

Home
Library
Facts
Speech recognition processing speed: number of sources

Fact · Speech recognition

7sources

The speech recognition processing speed dataset draws on 7 distinct sources.

Calculated from 66 data points on October 19, 2026

How it is calculated: Count of the distinct sources behind the dataset's 66 rows: 7.

A statistic of the dataset Speech Recognition Speed: Real-Time Factor Data, calculated by code from its published rows.

Calculated from

The best fine-tuned Whisper-small model, NOMIMOSE, processed audio at 5.99 times real time on an NVIDIA Jetson, while also having the lowest pooled word error rate of 22.72 percent.
On an NVIDIA Jetson, the pretrained Whisper-small model had a real-time factor of 0.212, which is 4.72 times faster than real time, on 264 cleft lip and palate utterances.
In the paper's Open ASR Leaderboard speed measurements on a single H100 GPU at batch size 1, the controlled autoregressive baseline ran at an RTFx of 12, meaning 12 times faster than real time.
In the paper's Open ASR Leaderboard speed measurements on a single H100 GPU at batch size 96, the controlled autoregressive baseline ran at an RTFx of 430, meaning 430 times faster than real time.
In the paper's Open ASR Leaderboard speed measurements on a single H100 GPU at batch size 1, the CTC model ran at an RTFx of 760, meaning 760 times faster than real time.
In the paper's Open ASR Leaderboard speed measurements on a single H100 GPU at batch size 96, the CTC model ran at an RTFx of 2584, meaning 2584 times faster than real time.
In the paper's Open ASR Leaderboard speed measurements on a single H100 GPU at batch size 1, NLE ran at an RTFx of 322, meaning 322 times faster than real time.
In the paper's Open ASR Leaderboard speed measurements on a single H100 GPU at batch size 1, NLE++ ran at an RTFx of 310, meaning 310 times faster than real time.
In the paper's Open ASR Leaderboard speed measurements on a single H100 GPU at batch size 96, NLE++ ran at an RTFx of 1630, meaning 1630 times faster than real time.
In the paper's Open ASR Leaderboard speed measurements on a single H100 GPU at batch size 96, NLE ran at an RTFx of 1722, meaning 1722 times faster than real time.
In the Galvez et al. 2024 paper, the baseline decoder for Parakeet RNN-T 1.1B ran at an RTFx of 414.63 on an A100-80GiB PCIe GPU at batch size 32, meaning 414.63 times faster than real time.
In the Galvez et al. 2024 paper, Parakeet RNN-T 1.1B with label looping and CUDA graphs ran at an RTFx of 1120.88 on an A100-80GiB PCIe GPU at batch size 32, meaning about 1,121 times faster than real time.
In our October 2026 benchmark on a Mac with an M2 Pro, Apple SpeechTranscriber transcribed 100 German FLEURS recordings about 72 times faster than real time.
In our October 2026 benchmark on a Mac with an M2 Pro, Apple SpeechTranscriber transcribed 100 English FLEURS recordings about 34 times faster than real time.
In our October 2026 benchmark on a Mac with an M2 Pro, Apple SpeechTranscriber transcribed 100 Spanish FLEURS recordings about 57 times faster than real time.
In our October 2026 benchmark on a Mac with an M2 Pro, Apple SpeechTranscriber transcribed 100 French FLEURS recordings about 46 times faster than real time.
In our October 2026 benchmark on a Mac with an M2 Pro, Apple SpeechTranscriber transcribed 100 Italian FLEURS recordings about 75 times faster than real time.
In our October 2026 benchmark on a Mac with an M2 Pro, Apple SpeechTranscriber transcribed 100 Portuguese FLEURS recordings about 51177 times faster than real time.
In our October 2026 benchmark on a Mac with an M2 Pro, Parakeet TDT 0.6B v3 transcribed 100 Bulgarian FLEURS recordings about 18 times faster than real time.
In our October 2026 benchmark on a Mac with an M2 Pro, Parakeet TDT 0.6B v3 transcribed 100 Czech FLEURS recordings about 19 times faster than real time.
In our October 2026 benchmark on a Mac with an M2 Pro, Parakeet TDT 0.6B v3 transcribed 100 Danish FLEURS recordings about 21 times faster than real time.
In our October 2026 benchmark on a Mac with an M2 Pro, Parakeet TDT 0.6B v3 transcribed 100 German FLEURS recordings about 20 times faster than real time.
In our October 2026 benchmark on a Mac with an M2 Pro, Parakeet TDT 0.6B v3 transcribed 100 Greek FLEURS recordings about 19 times faster than real time.
In our October 2026 benchmark on a Mac with an M2 Pro, Parakeet TDT 0.6B v3 transcribed 100 English FLEURS recordings about 19 times faster than real time.
In our October 2026 benchmark on a Mac with an M2 Pro, Parakeet TDT 0.6B v3 transcribed 100 Spanish FLEURS recordings about 22 times faster than real time.
In our October 2026 benchmark on a Mac with an M2 Pro, Parakeet TDT 0.6B v3 transcribed 100 Estonian FLEURS recordings about 22 times faster than real time.
In our October 2026 benchmark on a Mac with an M2 Pro, Parakeet TDT 0.6B v3 transcribed 100 Croatian FLEURS recordings about 20 times faster than real time.
In our October 2026 benchmark on a Mac with an M2 Pro, Parakeet TDT 0.6B v3 transcribed 100 Italian FLEURS recordings about 21 times faster than real time.
In our October 2026 benchmark on a Mac with an M2 Pro, Parakeet TDT 0.6B v3 transcribed 100 Lithuanian FLEURS recordings about 20 times faster than real time.
In our October 2026 benchmark on a Mac with an M2 Pro, Parakeet TDT 0.6B v3 transcribed 100 Maltese FLEURS recordings about 21 times faster than real time.
In our October 2026 benchmark on a Mac with an M2 Pro, Parakeet TDT 0.6B v3 transcribed 100 Dutch FLEURS recordings about 19 times faster than real time.
In our October 2026 benchmark on a Mac with an M2 Pro, Parakeet TDT 0.6B v3 transcribed 100 Polish FLEURS recordings about 19 times faster than real time.
In our October 2026 benchmark on a Mac with an M2 Pro, Parakeet TDT 0.6B v3 transcribed 100 Portuguese FLEURS recordings about 22 times faster than real time.
In our October 2026 benchmark on a Mac with an M2 Pro, Parakeet TDT 0.6B v3 transcribed 100 Russian FLEURS recordings about 21 times faster than real time.
In our October 2026 benchmark on a Mac with an M2 Pro, Parakeet TDT 0.6B v3 transcribed 100 Slovenian FLEURS recordings about 20 times faster than real time.
In our October 2026 benchmark on a Mac with an M2 Pro, Parakeet TDT 0.6B v3 transcribed 100 Swedish FLEURS recordings about 18 times faster than real time.
In our October 2026 benchmark on a Mac with an M2 Pro, Whisper large-v3-turbo transcribed 100 Bulgarian FLEURS recordings about 12 times faster than real time.
In our October 2026 benchmark on a Mac with an M2 Pro, Whisper large-v3-turbo transcribed 100 Czech FLEURS recordings about 14 times faster than real time.
In our October 2026 benchmark on a Mac with an M2 Pro, Whisper large-v3-turbo transcribed 100 Danish FLEURS recordings about 13 times faster than real time.
In our October 2026 benchmark on a Mac with an M2 Pro, Whisper large-v3-turbo transcribed 100 German FLEURS recordings about 16 times faster than real time.
In our October 2026 benchmark on a Mac with an M2 Pro, Whisper large-v3-turbo transcribed 100 Greek FLEURS recordings about 11 times faster than real time.
In our October 2026 benchmark on a Mac with an M2 Pro, Whisper large-v3-turbo transcribed 100 English FLEURS recordings about 12 times faster than real time.
In our October 2026 benchmark on a Mac with an M2 Pro, Whisper large-v3-turbo transcribed 100 Spanish FLEURS recordings about 15 times faster than real time.
In our October 2026 benchmark on a Mac with an M2 Pro, Whisper large-v3-turbo transcribed 100 Estonian FLEURS recordings about 15 times faster than real time.
In our October 2026 benchmark on a Mac with an M2 Pro, Whisper large-v3-turbo transcribed 100 French FLEURS recordings about 12 times faster than real time.
In our October 2026 benchmark on a Mac with an M2 Pro, Whisper large-v3-turbo transcribed 100 Italian FLEURS recordings about 17 times faster than real time.
In our October 2026 benchmark on a Mac with an M2 Pro, Whisper large-v3-turbo transcribed 100 Latvian FLEURS recordings about 15 times faster than real time.
In our October 2026 benchmark on a Mac with an M2 Pro, Whisper large-v3-turbo transcribed 100 Maltese FLEURS recordings about 13 times faster than real time.
In our October 2026 benchmark on a Mac with an M2 Pro, Whisper large-v3-turbo transcribed 100 Dutch FLEURS recordings about 12 times faster than real time.
In our October 2026 benchmark on a Mac with an M2 Pro, Whisper large-v3-turbo transcribed 100 Polish FLEURS recordings about 12 times faster than real time.
In our October 2026 benchmark on a Mac with an M2 Pro, Whisper large-v3-turbo transcribed 100 Portuguese FLEURS recordings about 16 times faster than real time.
In our October 2026 benchmark on a Mac with an M2 Pro, Whisper large-v3-turbo transcribed 100 Romanian FLEURS recordings about 12 times faster than real time.
In our October 2026 benchmark on a Mac with an M2 Pro, Whisper large-v3-turbo transcribed 100 Slovak FLEURS recordings about 14 times faster than real time.
In our October 2026 benchmark on a Mac with an M2 Pro, Whisper large-v3-turbo transcribed 100 Slovenian FLEURS recordings about 12 times faster than real time.
In the paper's Table 1, LSTM RNN-T model A (8 encoder layers of 640 cells) had a real-time factor of 0.53 when run on a single-core CPU of an Android device, about 1.89 times faster than real time.
In the paper's Table 1, LSTM RNN-T model B (encoder with 8x320 and 4x640 layers) had a real-time factor of 0.42 when run on a single-core CPU of an Android device, about 2.38 times faster than real time.
In the paper's Table 1, LSTM RNN-T model C (8 encoder layers of 640 cells, 4 predictor layers) had a real-time factor of 0.55 when run on a single-core CPU of an Android device, about 1.82 times faster than real time.
In the paper's Table 1, Emformer RNN-T model D had a real-time factor of 0.21 when run on a single-core CPU of an Android device, about 4.76 times faster than real time.
Smallest AI reports an effective real-time factor of about 1,000 for Pulse Pro on one NVIDIA L40S GPU when audio is sent in batched 50-second chunks without word timestamps.
Smallest AI reports that Pulse Pro transcribes a single two-hour audio file at about 200 times real time on one NVIDIA L40S GPU when word timestamps are on, in about 36 seconds.
Smallest AI reports that Pulse Pro transcribes a single two-hour audio file at 250 to 300 times real time on one NVIDIA L40S GPU when word timestamps are off, so about 275 times on average (midpoint).
In a test on AISHELL-1 run on a GeForce GTX TITAN X GPU, the spike-triggered non-autoregressive model at trigger threshold 0.1 had a real-time factor of 0.0059, about 169 times faster than real time.
In a test on AISHELL-1 run on a GeForce GTX TITAN X GPU, the spike-triggered non-autoregressive model at trigger threshold 0.3 had a real-time factor of 0.0056, about 179 times faster than real time.
In a test on AISHELL-1 run on a GeForce GTX TITAN X GPU, the spike-triggered non-autoregressive model at trigger threshold 0.5 had a real-time factor of 0.0055, about 182 times faster than real time.
In a test on AISHELL-1 run on a GeForce GTX TITAN X GPU, the spike-triggered non-autoregressive model at trigger threshold 0.7 had a real-time factor of 0.0054, about 185 times faster than real time.
In the MDM-ASR paper, the authors' system ran at an RTFx of 46.81 on LibriSpeech test-clean and test-other, measured on a single NVIDIA A100 GPU at batch size 1, which is 46.81 times faster than real time.

Cited on

What is the fastest speech-to-text API?Question

About Pricing Library Questions Glossary Data Market map Support Privacy Policy Terms of Service Get in touch

In your language:Eesti keel Latviešu valoda Lietuvių kalba Български Slovenščina Ελληνικά Українська Polski Čeština Slovenčina Magyar Română Hrvatski

Yap is a product of O-mega Enterprise Inc. © 2026

Made for thoughts that move faster than your fingers.

Speech recognition processing speed: number of sources | Yap