#2532 ValueError: Invalid input features shape: expected an input with shape (1, 80, 3000), but got an input with shape (1, 12

2402:800* Posted at: 1 day ago 👁13

Traceback (most recent call last):
File "videotrans\process\_overall.py", line 148, in run
File "faster_whisper\transcribe.py", line 1851, in restore_speech_timestamps
File "faster_whisper\transcribe.py", line 1190, in generate_segments
File "faster_whisper\transcribe.py", line 1400, in encode
ValueError: Invalid input features shape: expected an input with shape (1, 80, 3000), but got an input with shape (1, 128, 3000) instead

=
system:Windows-10-10.0.19045-SP0
version:v3.85
frozen:True
language:en
root_dir:C:/Users/Admin/Downloads/win-pyvideotrans-v3.85

1 Replies

Floor #1 DeepSeek Posted at 1 day ago

The error "Invalid input features shape: expected an input with shape (1, 80, 3000), but got an input with shape (1, 128, 3000)" indicates a mismatch between the audio features your file is producing and what the faster-whisper model expects. This is often caused by using a model that was trained on a different audio configuration (e.g., different number of Mel-frequency bands).

The most likely cause is that you are using the large-v3 model with an older version of faster-whisper that doesn't fully support it. The large-v3 model requires faster-whisper version 1.0.0 or higher.

Here is the step-by-step solution:

  1. Change your ASR model: In the main interface, under the "Speech Recognition (ASR)" section, change the model from large-v3 to large-v2 or medium. large-v2 is highly recommended for its balance of accuracy and speed.

  2. Clear the model cache (if the error persists):

    • Go to the software's installation directory (e.g., C:/Users/Admin/Downloads/win-pyvideotrans-v3.85).
    • Find and delete the models folder.
    • Restart the software. It will re-download the selected model (large-v2).
  3. If you must use large-v3: You need to ensure your faster-whisper library is up-to-date. Since you are using the pre-packaged Windows version, this is not directly user-upgradable. You would need to switch to the source code deployment method and run uv sync to get the latest dependencies, or wait for the software developer to release an updated package.

Immediate Action: Simply switching from large-v3 to large-v2 in the software's dropdown menu should resolve the issue immediately.

Please refer to the documentation for more information:

Post Your Reply
Open source and free maintenance is not easy. If this project is helpful to you, please consider making a small donation to help the project continue to maintain and update.

Donate: https://ko-fi.com/jianchang512

Trending Questions