How to know if it works
I am a total noob how do I know if it works it just doesn't print anything in the terminal, I use env btw
How did you run the code? You should use Python and the inference script. If you are unsure how to use it, you can paste the code into an AI chatbot and ask how to progress. Thanks for your interests.
How did you run the code? You should use Python and the inference script. If you are unsure how to use it, you can paste the code into an AI chatbot and ask how to progress. Thanks for your interests.
Yeah I will try to use inference script, I thought that is just for advanced use case lol. So then I will get kinda a timestamp file ? How can I make a translated transcription (subtitle, like a webvtt or srt file) out of them ?
You basically can't use this model alone to get a transcription. This is a VAD model that will only output the speech timestamps for an audio. However, you can reference https://github.com/TransWithAI/Faster-Whisper-TransWithAI-ChickenRice to find out how to use faster whisper and a proper whisper model to get transcriptions.
You basically can't use this model alone to get a transcription. This is a VAD model that will only output the speech timestamps for an audio. However, you can reference https://github.com/TransWithAI/Faster-Whisper-TransWithAI-ChickenRice to find out how to use faster whisper and a proper whisper model to get transcriptions.
Which model do you recommend ? I need a model for my rtx 4060 8gb, language is Japanese to English.
The default whisper from OpenAI will do translation to English, though it is not the best. Myself unfortunately do not have many recommendations on whisper. However, if you are interested in doing jp-en translation, we do have a currently free community product called Trans-In-Home that utilises multimodal ASR and other complicated model pipelines to do much better transcriptions and translation. We have a tg group at https://t.me/transWithAI for discussions around that.
Can I just tweak your Chickenrice model instead to make it translate to english ? I don't understand any Chinese so I can't discuss that but thanks
Unfortunately, the chickenrice one is only trained on the ja->cn corpus, so it has no generalisation capability. It is not possible to use the chickenrice model for English, you can still try the vanilla whisper though.
About that Trans-In-Home product, when will it release ?
It's already on and has been a trial platform for free for about 6 months. However, it's gonna end by 12/30, well, effectively tomorrow. I plan to make a full commitment from now on and start seeking VCs. If you'd still like a try now, I can DM you and give you the link. Though currently there is no comprehensive documentation on it, you may need to spend some time figuring out how to use it.
Please give me the link, I guess I have to educate myself more about this topic. I am pretty much a newbie so those are really new for me
Thank you. I managed to get the chickenrice model with faster-whisper running but there is this error, appreciate it if you can help me :
File "onnxruntime\capi\onnxruntime_inference_collection.py", line 266, in run
onnxruntime.capi.onnxruntime_pybind11_state.InvalidArgument: [ONNXRuntimeError] : 2 : INVALID_ARGUMENT : Got invalid dimensions for input: input_features for the following indices
index: 1 Got: 128 Expected: 80
Please fix either the inputs/outputs or the model.
[PYI-5296:ERROR] Failed to execute script 'infer' due to unhandled exception!