replicate / cog-whisper

Run OpenAI Whisper as a Cog model

Whisper Cog model

This is an implementation of Whisper as a Cog model. Cog packages machine learning models as standard containers.

First, run get_weights.sh from the project root to download pre-trained weights:

./scripts/get_weights.sh

You can then build a container and run predictions like so:

cog predict -i audio="<path/to/your/audio/file>"

cog-whisper's Issues

Cannot choose None in the Language option

The API reference states that we can specify None to detect language automatically. Doing this in Python returns this error:

replicate.exceptions.ReplicateError: - input.language: Invalid type. Expected: string, given: null

It appears that we are unable to choose None on the web demo of this as well. Please fix when you have time!

Upload big file

hello,
If the document is written, the upload file is limited to 25 MB, but when I tried it in replicate, I could upload up to 1 GB, and it ran, so I would like to know how it happened. thanks

translation to other languages

I often use translation, but there is onlya way to translate to English (boolean). It would be fantastic to have an option to translate to other language (polish in my case).

Tasks

Beta Give feedback

No tasks being tracked yet.

Python 3.10.4 - Prediction interrupted

Hello, I'm using your API for the past week and i started to get failures today when trying to use it.

Error:
raise ModelError(prediction.error)
replicate.exceptions.ModelError: Prediction interrupted; please retry (code: PA)

Full log:

Traceback (most recent call last):
File "/Users/x/.pyenv/versions/3.10.4/lib/python3.10/runpy.py", line 196, in _run_module_as_main
return _run_code(code, main_globals, None,
File "/Users/x/.pyenv/versions/3.10.4/lib/python3.10/runpy.py", line 86, in _run_code
exec(code, run_globals)
File "/Users/x/.vscode/extensions/ms-python.python-2023.10.0/pythonFiles/lib/python/debugpy/adapter/../../debugpy/launcher/../../debugpy/main.py", line 39, in
cli.main()
File "/Users/x/.vscode/extensions/ms-python.python-2023.10.0/pythonFiles/lib/python/debugpy/adapter/../../debugpy/launcher/../../debugpy/../debugpy/server/cli.py", line 430, in main
run()
File "/Users/x/.vscode/extensions/ms-python.python-2023.10.0/pythonFiles/lib/python/debugpy/adapter/../../debugpy/launcher/../../debugpy/../debugpy/server/cli.py", line 284, in run_file
runpy.run_path(target, run_name="main")
File "/Users/x/.vscode/extensions/ms-python.python-2023.10.0/pythonFiles/lib/python/debugpy/_vendored/pydevd/_pydevd_bundle/pydevd_runpy.py", line 321, in run_path
return _run_module_code(code, init_globals, run_name,
File "/Users/x/.vscode/extensions/ms-python.python-2023.10.0/pythonFiles/lib/python/debugpy/_vendored/pydevd/_pydevd_bundle/pydevd_runpy.py", line 135, in _run_module_code
_run_code(code, mod_globals, init_globals,
File "/Users/x/.vscode/extensions/ms-python.python-2023.10.0/pythonFiles/lib/python/debugpy/_vendored/pydevd/_pydevd_bundle/pydevd_runpy.py", line 124, in _run_code
exec(code, run_globals)
File "/Users/x/Desktop/Work/xy/app/test.py", line 48, in
main()
File "/Users/x/Desktop/Work/xy/app/test.py", line 37, in main
output = replicate.run(
File "/Users/x/.pyenv/versions/3.10.4/lib/python3.10/site-packages/replicate/client.py", line 140, in run

This is quite urgent

"Output Not found" error on replicate.com

Same results whether signed in or using public demo.

Is this the right place to report this problem? Sorry in advance if not.

Languages keys are not properly linked to their ISO standards

It seems all if not most languages in the language selection field are not linked to their ISO standard.
For example selecting "el" for Greek will work, but selecting "Greek" itself will throw a key not found error.
Should be easy to fix.

Doesn't work if title contains #

When I upload and start the model

Error : Prediction failed.
Unknown error handling prediction.

Faster responses?

Why not using whisper jax for faster responses?

Incompatible Error: GPU false + MacBook Pro M2: Cog + Whisper

Incompatible Error: GPU false + MacBook Pro M2: Cog + Whisper
replicate/cog#1179

add word timestamps from upstream

Hi

I noticed the official OpenAI whisper includes support for word timestamps.

There is a demande for this feature, as can be seen from the 40k+ uses of this replicate model

Can you consider exposing the args for this feature on your replicate model?

Thank you.