seamless_communication

seamless_communication copied to clipboard

Does speech recognition support streaming reasoning? How can I change it?

RyanFlying

Translation result is not good.

1

Regarding the following sentence, I think the translation result is not as good as Google Translate. ``` Hi, it's been fans, welcome to another installment at Spring Tips. ``` The...

asasas234

Finetuning TTS, training crashes after eval

4

I try to finetune model for TTS. I have myself prepared train_manifest.json and eval_manifest.json files with coding instead of using the datasets scripts. My train_manifest.json should look like the same...

R4ZZ3

When I run the code below, there is a warning message. Is it an error? How can I handle it? ```python from transformers import SeamlessM4TTokenizer tokenizer = SeamlessM4TTokenizer.from_pretrained( "facebook/hf-seamless-m4t-medium", src_lang="eng"...

leobavila

Why such a simple example is wrong?

2

```python # T2ST input_text = "how do you do" src_lang = "eng" tgt_lang = "eng" path_to_save_audio = "./test.wav" translated_text, wav, sr = translator.predict(input_text, "t2st", tgt_lang, src_lang=src_lang, ngram_filtering=True) # print(wav.shape) torchaudio.save(path_to_save_audio,...

HLearning

The installation cannot be competed

39

PIP version: `pip 23.2.1` Python version: `python 3.10` Error message after running `pip install fairseq2==0.1`: ``` Collecting fairseq2==0.1 Obtaining dependency information for fairseq2==0.1 from https://files.pythonhosted.org/packages/cd/27/46c14e28e8cb0aa602660ce64d4547a37f460d382e4fcf94f2a53d47e5b0/fairseq2-0.1.0-py3-none-any.whl.metadata Using cached fairseq2-0.1.0-py3-none-any.whl.metadata (1.2 kB)...

AlekseiLitvin

Added pyproject.toml to lock down dependency versions

1

## Summary - added pyproject.toml file for building the library, following [PEP 518](https://peps.python.org/pep-0518/); - added poetry-based dependency management to lock down dependency versions for reproducibility; ## Tests - build successfully...

jasonmusespresso

Error when reconstructing aligned data

7

When I use the `wet_lines` script to download and gather aligned text information from the metadata, there is something wrong. The error message is as below. So what should I...

starshine360

Audio is truncated in the output file

4

I have noticed that when the input audio file exceeds approximately 30 seconds in duration, the resulting output file contains only the first 10 seconds or the last 10 seconds...

tomchang25

if it has how to use it. translated_text, wav, sr = translator.predict( input='/content/drive/MyDrive/GPI/WAV/1.wav', task_str='s2st', tgt_lang='spa', # target language src_lang='spa', # source language # If you specify this, it will improve...

Nachuwu