Fact · Speech recognition
0.0055real-time factor
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.
Quote found word for word on the live source, last checked October 19, 2026
In the source's own words
“on a GeForce GTX TITAN X 12G GPU. ... Threshold β 0.1 0.3 0.5 0.7 Performance 7.88 7.67 7.77 7.70 Seconds 212.04 202.59 200.62 198.44 RTF 0.0059 0.0056 0.0055 0.0054”
Spike-Triggered Non-Autoregressive Transformer for End-to-End Speech Recognition, Zhengkun Tian et al., Interspeech 2020, 2020, Section 3.3, Table 2.
Open the original (DOI 10.21437/Interspeech.2020-2086)- What it measures
- Table 2 of the paper records the time ST-NAT spends decoding the test set at four trigger thresholds, and the real-time factor is that time divided by the audio length. Utterances were decoded one by one on a single GPU.
- What it does not show
- Measured on a 2020 GPU with one decoding setup on AISHELL-1, so the speed is a relative figure and not a product benchmark. The paper's own RTF has two significant digits, so the converted speed is approximate.
In the data
A row of Speech Recognition Speed: Real-Time Factor Data, as Tian et al. 2020, threshold 0.5: 182 times faster than real time.
- System
- ST-NAT (trigger threshold 0.5)
- Kind
- Research system
- Hardware
- GeForce GTX TITAN X GPU
- Test set
- AISHELL-1 test set
- Year
- 2020
- Note
- 1 divided by the real-time factor 0.0055 is 181.82 times faster than real time.