← All work
Qwen3-ASR Service
Qwen3-ASR-1.7B on RunPod serverless GPUs: multilingual speech recognition with context biasing, language identification, and word-level timestamps.
The problem
Speech recognition is bursty work. Keeping a GPU warm around the clock for it wastes most of what you pay for.
What I built
- Qwen3-ASR-1.7B packaged as a RunPod serverless worker, so a GPU is only billed while a request is running.
- Context biasing, language identification, and word-level timestamps exposed through the API.
- Published on the RunPod Hub so it can be deployed by anyone in a couple of clicks.
What it moved
- Listed on the RunPod Hub with the source public on GitHub.