Local Speech-to-Text — Faster Whisper v0.2 (Free Open Source, CPU)



📚Video Course
🕒Last updated: 08/19/2024
👥Entrepreneurs / Self-media creators / Cross-border e-commerce sellers / Side-hustlers

Study material — included with membership
Unlock Course with Membership
🔓
Get this course for reference study with membership
📦
Join membership — study all 16742 courses & software tools as reference
🔄
Courses keep updating — members study new material as it arrives
Share:

Recently I needed to convert a recording to text. After searching, most online tools are paid, and free ones rely on cloud APIs or require a CUDA GPU. So I decided to build a small speech-to-text tool based on the Faster Whisper model that achieves great results even on an ordinary CPU.

\ \

\ \

How to use

\ \
    \ \
  1. Launch the program:Double-click the file to open the tool.
  2. \ \
  3. Select a model:Choose the model you want to use in the top-left corner.
  4. \ \
  5. Set inference method and precision:Choose the inference method and precision in the bottom-left corner. On CPU, it is recommended to select int8 mode, which is faster.
  6. \ \
  7. Note on GPU selection:If your PC lacks a CUDA GPU, don't select the GPU option, or the program may crash. We planned to detect CUDA via PyTorch, but that made the package too large, so we ultimately skipped that approach.
  8. \ \
  9. downloadmodel:The model must be downloaded on first use. You can view the download progress in the console (console), please wait patiently.
  10. \ \
\ \

Updates and FAQ

\ \
\ \
1. Compatibility issues:
\ \
This tool requires 64-bit Windows 10 or later; it doesn't support older Windows 7, XP and similar.
\ \
Some low-end Pentium and Celeron processors may fail to run properly due to missing required instruction sets.
\ \
2. Model selection:
\ \
Use large model, it is recommended to set the precision to float32。
\ \
For the other two models, use int8 precision.
\ \
3. Network issues:
\ \
The model files are large, please ensure a stable network connection. If you encounter download failures, it is recommended to delete the models folder in the same directory, then retry the download.
\ \
\ \

If the above didn't solve it, comment with your feedback and the console output so we can help better. Thanks!

\ \

 

\ \ \ \

Related Courses

Email