This project evaluates various Speech-to-Text (STT) configurations for both streaming and batch (offline) transcription, testing different model variants, container environments, model serving frameworks, deployment platforms, and hardware configurations.
OpenAI Whisper, Faster-Whisper, NVIDIA NeMo ASR, and Wav2Vec are all speech-to-text (ASR) models, but they differ in terms of architecture, performance, hardware requirements, and use cases. Here’s how they compare:
| OpenAI Whisper | Faster-Whisper | NeMo | |
|---|---|---|---|
| Developer | OpenAI | OpenAI (optimized by Faster-Whisper) | |
| Model Type | Transformer-based | Transformer-based | Conformer-based |
| Pretraining Method | Supervised (large-scale labeled data) | Optimized Whisper | Supervised & Semi-Supervised |
| Multilingual | Yes (99+ languages) | Yes (99+ languages) | Yes (for some models) |
| Hardware | CPU/GPU | Optimized for GPU | Optimized for NVIDIA GPUs |
| Streaming | No | No | Yes |
Tracking the deployment and integration status of different speech-to-text (ASR) models (OpenAI Whisper, Faster-Whisper, NVIDIA NeMo ASR, and Wav2Vec 2.0) across various environments, model servers, and performance metrics.
| OpenAI Whisper | Faster-Whisper | NeMo | |
|---|---|---|---|
| Ubuntu Dockerfile | X | TODO | TODO |
| UBI Dockerfile | X | TODO | TODO |
| ModelKit | TODO | TODO | TODO |
| RHEL OS | X | TODO | TODO |
| OCP | X | TODO | TODO |
| Embedded Server | X | TODO | TODO |
| Built-in Server | X | TODO | TODO |
| Decoupled Server | TODO | TODO | TODO |
| NVIDIA Triton | TODO | TODO | TODO |
| vLLM | TODO | TODO | TODO |
| Ray Serve | TODO | TODO | TODO |
| Batch | X | TODO | TODO |
| Streaming | TODO | TODO | TODO |
| Word Error Rate (WER) | X | TODO | TODO |
| Match Error Rate (MER) | X | TODO | TODO |
| Word Information Lost (WIL) | X | TODO | TODO |
| Word Information Preserved (WIP) | X | TODO | TODO |
| Character Error Rate (CER) | X | TODO | TODO |
| Pipeline Build | TODO | TODO | TODO |
| Summary | TODO | TODO | TODO |
-
Request environments from demo.redhat.com
-
- Activity:
Practice / Enablement - Purpose:
Trying out a technical solution - Region:
us-east-2 - GPU Selection by Node Type:
g6.xlarge 1 x L4ORg6.12xlarge 4 x L4
- Activity:
-
AWS with OpenShift Open Environment
- Activity:
Practice / Enablement - Purpose:
Trying out a technical solution - Region:
us-east-2 - OpenShift Version:
4.17 - Control Plane Count:
1 - Control Plane Instance Type:
m6a.4xlarge
- Activity:
-
- Activity:
Practice / Enablement - Purpose:
Trying out a technical solution - Developer Hub:
stable - Orchestrator:
Enabled - Enable RHOAI:
Enabled
- Activity:
Go to the Crawl README
TODO
Go to the Run README