Whisper-Tiny: Optimized for Qualcomm Devices
HuggingFace Whisper-Tiny ASR (Automatic Speech Recognition) model is a state-of-the-art system designed for transcribing spoken language into written text. This model is based on the transformer architecture and has been optimized for edge inference by replacing Multi-Head Attention (MHA) with Single-Head Attention (SHA) and linear layers with convolutional (conv) layers. It exhibits robust performance in realistic, noisy environments, making it highly reliable for real-world applications. Specifically, it excels in long-form transcription, capable of accurately transcribing audio clips up to 30 seconds long. Time to the first token is the encoder's latency, while time to each additional token is decoder's latency, where we assume a max decoded length specified below.
This is based on the implementation of Whisper-Tiny found here. This repository contains pre-exported model files optimized for Qualcomm® devices. You can use the Qualcomm® AI Hub Models library to export with custom configurations. More details on model performance across various devices, can be found here.
Qualcomm AI Hub Models uses Qualcomm AI Hub Workbench to compile, profile, and evaluate this model. Sign up to run these models on a hosted Qualcomm® device.
Deploying Whisper-Tiny on-device
This model is compatible with the Qualcomm Voice AI SDK. Download the SDK from the Qualcomm Package Manager to deploy this model on-device.
Getting Started
There are two ways to deploy this model on your device:
Option 1: Download Pre-Exported Models
Below are pre-exported model assets ready for deployment.
| Runtime | Precision | Chipset | SDK Versions | Download |
|---|---|---|---|---|
| PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | QAIRT 2.50, ONNX Runtime 1.30.0 | Download |
| PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | QAIRT 2.50, ONNX Runtime 1.30.0 | Download |
| PRECOMPILED_QNN_ONNX | float | Snapdragon® X2 Elite | QAIRT 2.50, ONNX Runtime 1.30.0 | Download |
| PRECOMPILED_QNN_ONNX | float | Snapdragon® X Elite | QAIRT 2.50, ONNX Runtime 1.30.0 | Download |
| PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Gen 3 Mobile | QAIRT 2.50, ONNX Runtime 1.30.0 | Download |
| PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Gen 1 Mobile | QAIRT 2.50, ONNX Runtime 1.30.0 | Download |
| PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ IQ-8275 | QAIRT 2.50, ONNX Runtime 1.30.0 | Download |
| PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | QAIRT 2.50, ONNX Runtime 1.30.0 | Download |
| PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | QAIRT 2.50, ONNX Runtime 1.30.0 | Download |
| QNN_CONTEXT_BINARY | float | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | QAIRT 2.50 | Download |
| QNN_CONTEXT_BINARY | float | Snapdragon® 8 Elite For Galaxy Mobile | QAIRT 2.50 | Download |
| QNN_CONTEXT_BINARY | float | Snapdragon® X2 Elite | QAIRT 2.50 | Download |
| QNN_CONTEXT_BINARY | float | Snapdragon® X Elite | QAIRT 2.50 | Download |
| QNN_CONTEXT_BINARY | float | Snapdragon® 8 Gen 3 Mobile | QAIRT 2.50 | Download |
| QNN_CONTEXT_BINARY | float | Snapdragon® 8 Gen 1 Mobile | QAIRT 2.50 | Download |
| QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ IQ-8275 | QAIRT 2.50 | Download |
| QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | QAIRT 2.50 | Download |
| QNN_CONTEXT_BINARY | float | Qualcomm® SA8775P | QAIRT 2.50 | Download |
| QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ IQ-9075 | QAIRT 2.50 | Download |
| QNN_CONTEXT_BINARY | float | Qualcomm® SA7255P | QAIRT 2.50 | Download |
| QNN_CONTEXT_BINARY | float | Qualcomm® SA8295P | QAIRT 2.50 | Download |
| VOICE_AI | float | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | QAIRT 2.50 | Download |
| VOICE_AI | float | Snapdragon® 8 Elite For Galaxy Mobile | QAIRT 2.50 | Download |
| VOICE_AI | float | Snapdragon® X2 Elite | QAIRT 2.50 | Download |
| VOICE_AI | float | Snapdragon® X Elite | QAIRT 2.50 | Download |
| VOICE_AI | float | Snapdragon® 8 Gen 3 Mobile | QAIRT 2.50 | Download |
| VOICE_AI | float | Snapdragon® 8 Gen 1 Mobile | QAIRT 2.50 | Download |
| VOICE_AI | float | Qualcomm® Dragonwing™ IQ-8275 | QAIRT 2.50 | Download |
| VOICE_AI | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | QAIRT 2.50 | Download |
| VOICE_AI | float | Qualcomm® SA8775P | QAIRT 2.50 | Download |
| VOICE_AI | float | Qualcomm® Dragonwing™ IQ-9075 | QAIRT 2.50 | Download |
| VOICE_AI | float | Qualcomm® SA7255P | QAIRT 2.50 | Download |
| VOICE_AI | float | Qualcomm® SA8295P | QAIRT 2.50 | Download |
For more device-specific assets and performance metrics, visit Whisper-Tiny on Qualcomm® AI Hub.
Option 2: Export with Custom Configurations
Use the Qualcomm® AI Hub Models Python library to compile and export the model with your own:
- Custom weights (e.g., fine-tuned checkpoints)
- Custom input shapes
- Target device and runtime configurations
This option is ideal if you need to customize the model beyond the default configuration provided here.
See our repository for Whisper-Tiny on GitHub for usage instructions.
Model Details
Model Type: Model_use_case.speech_recognition
Model Stats:
- Input resolution: 80x3000 (30 seconds audio)
- Max decoded sequence length: 200 tokens
- Model checkpoint: openai/whisper-tiny
- Model size (decoder) (float): 109 MB
- Model size (encoder) (float): 35.9 MB
- Number of parameters (decoder): 28.4M
- Number of parameters (encoder): 9.39M
Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit |
|---|---|---|---|---|---|---|
| decoder | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 1.515 ms | 0 - 7 MB | NPU |
| decoder | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 1.555 ms | 0 - 6 MB | NPU |
| decoder | PRECOMPILED_QNN_ONNX | float | Snapdragon® X2 Elite | 1.418 ms | 10 - 10 MB | NPU |
| decoder | PRECOMPILED_QNN_ONNX | float | Snapdragon® X Elite | 2.201 ms | 84 - 84 MB | NPU |
| decoder | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Gen 3 Mobile | 1.962 ms | 12 - 19 MB | NPU |
| decoder | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Gen 1 Mobile | 2.632 ms | 13 - 28 MB | NPU |
| decoder | PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ IQ-8275 | 3.362 ms | 10 - 24 MB | NPU |
| decoder | PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 2.484 ms | 0 - 87 MB | NPU |
| decoder | PRECOMPILED_QNN_ONNX | float | Qualcomm® QCS8450 | 2.632 ms | 13 - 28 MB | NPU |
| decoder | PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 3.184 ms | 10 - 23 MB | NPU |
| decoder | PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 2.201 ms | 84 - 84 MB | NPU |
| decoder | PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 1.555 ms | 0 - 6 MB | NPU |
| decoder | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 1.547 ms | 0 - 9 MB | NPU |
| decoder | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Elite For Galaxy Mobile | 1.613 ms | 0 - 12 MB | NPU |
| decoder | QNN_CONTEXT_BINARY | float | Snapdragon® X2 Elite | 1.77 ms | 10 - 10 MB | NPU |
| decoder | QNN_CONTEXT_BINARY | float | Snapdragon® X Elite | 2.435 ms | 10 - 10 MB | NPU |
| decoder | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Gen 3 Mobile | 1.873 ms | 0 - 7 MB | NPU |
| decoder | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Gen 1 Mobile | 2.545 ms | 2 - 12 MB | NPU |
| decoder | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ IQ-8275 | 3.219 ms | 10 - 24 MB | NPU |
| decoder | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 2.426 ms | 10 - 12 MB | NPU |
| decoder | QNN_CONTEXT_BINARY | float | Qualcomm® SA8650P | 3.164 ms | 10 - 19 MB | NPU |
| decoder | QNN_CONTEXT_BINARY | float | Qualcomm® SA8255P | 3.164 ms | 10 - 19 MB | NPU |
| decoder | QNN_CONTEXT_BINARY | float | Qualcomm® QCS8450 | 2.545 ms | 2 - 12 MB | NPU |
| decoder | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ IQ-9075 | 3.067 ms | 10 - 23 MB | NPU |
| decoder | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ IQ-X7181 | 2.435 ms | 10 - 10 MB | NPU |
| decoder | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ Q-8750 | 1.613 ms | 0 - 12 MB | NPU |
| decoder | QNN_CONTEXT_BINARY | float | Qualcomm® SA8295P | 2.836 ms | 10 - 15 MB | NPU |
| decoder | VOICE_AI | float | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 1.523 ms | 0 - 8 MB | NPU |
| decoder | VOICE_AI | float | Snapdragon® 8 Elite For Galaxy Mobile | 1.552 ms | 0 - 8 MB | NPU |
| decoder | VOICE_AI | float | Snapdragon® X2 Elite | 1.778 ms | 10 - 10 MB | NPU |
| decoder | VOICE_AI | float | Snapdragon® X Elite | 2.444 ms | 10 - 10 MB | NPU |
| decoder | VOICE_AI | float | Snapdragon® 8 Gen 3 Mobile | 1.875 ms | 0 - 8 MB | NPU |
| decoder | VOICE_AI | float | Snapdragon® 8 Gen 1 Mobile | 2.529 ms | 1 - 11 MB | NPU |
| decoder | VOICE_AI | float | Qualcomm® Dragonwing™ IQ-8275 | 3.216 ms | 10 - 24 MB | NPU |
| decoder | VOICE_AI | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 2.416 ms | 10 - 12 MB | NPU |
| decoder | VOICE_AI | float | Qualcomm® SA8650P | 3.236 ms | 5 - 14 MB | NPU |
| decoder | VOICE_AI | float | Qualcomm® SA8255P | 3.236 ms | 5 - 14 MB | NPU |
| decoder | VOICE_AI | float | Qualcomm® QCS8450 | 2.529 ms | 1 - 11 MB | NPU |
| decoder | VOICE_AI | float | Qualcomm® Dragonwing™ IQ-9075 | 3.061 ms | 10 - 23 MB | NPU |
| decoder | VOICE_AI | float | Qualcomm® Dragonwing™ IQ-X7181 | 2.444 ms | 10 - 10 MB | NPU |
| decoder | VOICE_AI | float | Qualcomm® Dragonwing™ Q-8750 | 1.552 ms | 0 - 8 MB | NPU |
| decoder | VOICE_AI | float | Qualcomm® SA8295P | 2.834 ms | 10 - 15 MB | NPU |
| encoder | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 11.842 ms | 21 - 29 MB | NPU |
| encoder | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 14.215 ms | 20 - 26 MB | NPU |
| encoder | PRECOMPILED_QNN_ONNX | float | Snapdragon® X2 Elite | 11.674 ms | 33 - 33 MB | NPU |
| encoder | PRECOMPILED_QNN_ONNX | float | Snapdragon® X Elite | 24.559 ms | 33 - 33 MB | NPU |
| encoder | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Gen 3 Mobile | 17.981 ms | 20 - 28 MB | NPU |
| encoder | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Gen 1 Mobile | 35.437 ms | 19 - 34 MB | NPU |
| encoder | PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ IQ-8275 | 27.607 ms | 19 - 23 MB | NPU |
| encoder | PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 24.189 ms | 12 - 33 MB | NPU |
| encoder | PRECOMPILED_QNN_ONNX | float | Qualcomm® QCS8450 | 35.437 ms | 19 - 34 MB | NPU |
| encoder | PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 27.57 ms | 23 - 26 MB | NPU |
| encoder | PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 24.559 ms | 33 - 33 MB | NPU |
| encoder | PRECOMPILED_QNN_ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 14.215 ms | 20 - 26 MB | NPU |
| encoder | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 11.853 ms | 0 - 7 MB | NPU |
| encoder | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Elite For Galaxy Mobile | 14.14 ms | 1 - 9 MB | NPU |
| encoder | QNN_CONTEXT_BINARY | float | Snapdragon® X2 Elite | 12.263 ms | 0 - 0 MB | NPU |
| encoder | QNN_CONTEXT_BINARY | float | Snapdragon® X Elite | 24.871 ms | 0 - 0 MB | NPU |
| encoder | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Gen 3 Mobile | 17.829 ms | 1 - 8 MB | NPU |
| encoder | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Gen 1 Mobile | 34.718 ms | 0 - 10 MB | NPU |
| encoder | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ IQ-8275 | 27.57 ms | 0 - 13 MB | NPU |
| encoder | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 23.837 ms | 1 - 4 MB | NPU |
| encoder | QNN_CONTEXT_BINARY | float | Qualcomm® SA8650P | 28.644 ms | 1 - 10 MB | NPU |
| encoder | QNN_CONTEXT_BINARY | float | Qualcomm® SA8255P | 28.644 ms | 1 - 10 MB | NPU |
| encoder | QNN_CONTEXT_BINARY | float | Qualcomm® QCS8450 | 34.718 ms | 0 - 10 MB | NPU |
| encoder | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ IQ-9075 | 27.334 ms | 2 - 14 MB | NPU |
| encoder | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ IQ-X7181 | 24.871 ms | 0 - 0 MB | NPU |
| encoder | QNN_CONTEXT_BINARY | float | Qualcomm® Dragonwing™ Q-8750 | 14.14 ms | 1 - 9 MB | NPU |
| encoder | QNN_CONTEXT_BINARY | float | Qualcomm® SA8295P | 34.397 ms | 1 - 6 MB | NPU |
| encoder | VOICE_AI | float | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 11.854 ms | 0 - 8 MB | NPU |
| encoder | VOICE_AI | float | Snapdragon® 8 Elite For Galaxy Mobile | 14.08 ms | 1 - 9 MB | NPU |
| encoder | VOICE_AI | float | Snapdragon® X2 Elite | 12.044 ms | 0 - 0 MB | NPU |
| encoder | VOICE_AI | float | Snapdragon® X Elite | 25.022 ms | 0 - 0 MB | NPU |
| encoder | VOICE_AI | float | Snapdragon® 8 Gen 3 Mobile | 17.942 ms | 1 - 9 MB | NPU |
| encoder | VOICE_AI | float | Snapdragon® 8 Gen 1 Mobile | 34.609 ms | 1 - 11 MB | NPU |
| encoder | VOICE_AI | float | Qualcomm® Dragonwing™ IQ-8275 | 27.634 ms | 0 - 13 MB | NPU |
| encoder | VOICE_AI | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 23.767 ms | 0 - 2 MB | NPU |
| encoder | VOICE_AI | float | Qualcomm® SA8650P | 28.668 ms | 0 - 6 MB | NPU |
| encoder | VOICE_AI | float | Qualcomm® SA8255P | 28.668 ms | 0 - 6 MB | NPU |
| encoder | VOICE_AI | float | Qualcomm® QCS8450 | 34.609 ms | 1 - 11 MB | NPU |
| encoder | VOICE_AI | float | Qualcomm® Dragonwing™ IQ-9075 | 27.413 ms | 2 - 14 MB | NPU |
| encoder | VOICE_AI | float | Qualcomm® Dragonwing™ IQ-X7181 | 25.022 ms | 0 - 0 MB | NPU |
| encoder | VOICE_AI | float | Qualcomm® Dragonwing™ Q-8750 | 14.08 ms | 1 - 9 MB | NPU |
| encoder | VOICE_AI | float | Qualcomm® SA8295P | 34.377 ms | 0 - 6 MB | NPU |
License
- The license for the original implementation of Whisper-Tiny can be found here.
References
Community
- Join our AI Hub Slack community to collaborate, post questions and learn more about on-device AI.
- For questions or feedback please reach out to us.
