Marksdo / Whisper

Batch Local Transcribe Audio/Movie To Text With Whisper AI Model. Keep Privacy Safe!
https://whisper.marksdo.com
19 stars 4 forks source link

Problema al transcribir Japones #38

Open Jack-479 opened 3 months ago

Jack-479 commented 3 months ago

no genera tiempos de subtítulos, genera un archivo de texto plano, ademas al realizar una transcripción de archivo de 01 horas a 01:30 lo realiza en un tiempo de 5 a 6 horas aprox. utilizando un modelo pequeño inclusive, lo hago en un Mac air M1.

al ejecutar Deepgram si bien es muy rápido, el texto transcrito viene lleno de errores de ortografía y gramaticales

Marksdo commented 3 months ago

Here are some settings to improve transcription speed:

1. Use Medium and Small model types.

2. Use fine-tuned Japanese language models (you may need to switch between several fine-tuned models to find the one that best suits your speech scenario).

image

3. Enable the GPU acceleration option for Apple Silicon machines.

image

4. Use the Fast template for the Extra parameter.

image

5. Use the Quick Cut feature to split long videos into smaller segments of less than 30 minutes (Cut on silent piece), and select chunk-based transcription in the strategy module.

image image

6. Plug in the laptop's power.