requests==2.28.1 flask==3.0.2 pydub pydantic nump...

The error you're seeing is caused by Python version incompatibility with some packages in your requirements list, most notably LangSegment>=0.2.6.

---

🧾 Problem Summary

You're getting messages like:

text
Ignored the following versions that require a different python version: ...
Could not find a version that satisfies the requirement LangSegment>=0.2.6

This means:

1. Some packages you listed don't support your current Python version.
2. Specifically, LangSegment>=0.2.6 does not exist on PyPI, at least not for your Python version.

---

🔍 Root Cause

According to PyPI:

• The latest available version of LangSegment is 0.2.0.
• Version 0.2.6 or higher does not exist (as of now).
• The error about ignored versions suggests you're using a Python version outside of the range these packages support, possibly Python 3.10 or newer.

---

✅ Solutions

1. Fix the LangSegment Version

Replace:

text
LangSegment>=0.2.6

With:

text
LangSegment==0.2.0

This will stop pip from trying to fetch a non-existent version.

---

2. Use a Compatible Python Version

Some packages (like older funasr or LangSegment) might need:

• Python 3.7 - 3.9

Try setting up a virtual environment with a compatible Python version:

bash
# Using pyenv for example
pyenv install 3.8.18
pyenv virtualenv 3.8.18 langenv
pyenv activate langenv
pip install -r requirements.txt

---

3. Identify All Python-Version-Restricted Packages

To catch more of these errors early:

bash
pip install pipdeptree
pipdeptree --warn silence | grep -i requires-python

---

🧹 Suggested Cleaned requirements.txt

Here’s an edited version of your file that should work under Python 3.8–3.9:

text
requests==2.28.1
flask==3.0.2
pydub
pydantic
numpy
scipy
tensorboard
librosa==0.9.2
numba
pytorch-lightning
gradio==4.19
gradio_client
ffmpeg-python
onnxruntime
tqdm
funasr==1.0.0
cn2an
pypinyin
pyopenjtalk
g2p_en
torchaudio
modelscope==1.10.0
sentencepiece
transformers
chardet
PyYAML
psutil
jieba_fast
jieba
LangSegment==0.2.0
Faster_Whisper
flask_httpauth
flask_cors

Let me know your exact Python version if you'd like help tailoring the list more precisely.

Below is a concrete game‑plan to eliminate the ImportError: cannot import name 'setLangfilters' that stops GPT_SoVITS/TTS_infer_pack/TextPreprocessor.py on Python 3.10.

---

Key takeaway (one‑paragraph summary)

LangSegment 0.2.0 is the only version currently on PyPI, but the GPT‑SoVITS v2 code you are running was written for the now‑vanished LangSegment 0.3.5+, which contained the functions setLangfilters/getLangfilters. Because those symbols were dropped when the maintainer yanked anything newer than 0.2.0, importing fails on every modern install. You can solve the problem in three main ways: (1) pin and install a community‑archived 0.3.5 wheel; (2) patch GPT‑SoVITS to stop calling the missing functions; or (3) upgrade to GPT‑SoVITS v3+, which removed the LangSegment dependency altogether. Most users choose option 1 because it is quickest; option 3 is the cleanest long‑term; option 2 is a fallback if corporate policy forbids third‑party wheels.

---

1 What happened?

Fact	Evidence
Only LangSegment 0.2.0 is on PyPI today
GPT‑SoVITS v2 depends on setLangfilters, present in ≥0.3.5
0.3.5 was yanked; several projects now hit the same ImportError	GitHub GitHub
A backup of 0.3.5 exists on GitHub and still works
0.3.5 can still be installed through a direct wheel or git URL	GitHub
GPT‑SoVITS v3 migrated away from LangSegment

---

2 Quick‑fix: install the archived 0.3.5 build

Works in minutes; best for getting your current code running right now.

bash
# Create an isolated env (Python 3.8 or 3.9 is safest)
conda create -n sovits38 python=3.8 -y
conda activate sovits38

# Option A – install from the community backup repo
pip install git+https://github.com/chameleon-ai/LangSegment-0.3.5-backup.git@main  # 

# Option B – grab the wheel cached by Libraries.io
pip install LangSegment==0.3.5 --index-url https://pypi.python.org/simple --extra-index-url=https://huggingface.github.io/mirror/  # wheel still mirrored here 

Why Python 3.8/3.9? 0.3.5 was released before Python 3.10; some users have reported minor warnings on 3.10+ but it runs fine in 3.8/3.9 GitHub.

After the install, run:

bash
python -c "import LangSegment, inspect, sys; print(LangSegment.setLangfilters, sys.version)"

If you see a function reference and no traceback, the library is healthy.

---

3 Long‑term options

3.1 Upgrade your project to GPT‑SoVITS v3+

• v3 completely drops LangSegment and uses a lighter‑weight tokenizer .
• You only have to retrain the text‑preprocessor; models themselves still load.

3.2 Patch the code instead of the library

If you must stay on 0.2.0, monkey‑patch missing APIs right after the import:

python
import LangSegment
if not hasattr(LangSegment, "setLangfilters"):
    def _noop(*args, **kwargs): pass
    LangSegment.setLangfilters = _noop
    LangSegment.getLangfilters = _noop

Several projects do exactly this as a hot‑fix GitHub.
Down‑side: you lose the multilingual filter feature, so mixed‑language text may segment less cleanly.

---

4 Sanity‑check your requirements.txt

1. Pin the rescued version so new teammates don’t break their env:

   text
   # Language segmentation
   LangSegment==0.3.5  # pinned – 0.2.0 lacks setLangfilters

2. Add a Python version marker if you keep using 3.8/3.9:

   text
   python_version < "3.10"

3. Run pip-compile --generate-hashes (if you use pip‑tools) so future builds are reproducible.

---

5 Troubleshooting matrix

Symptom	Fix
pip still fetches 0.2.0	Clear wheel cache pip cache purge; double‑check the --extra-index-url.
Wheel won’t install on 3.10+	Switch to 3.8/3.9 or build from source with python -m pip install --no-binary :all: LangSegment==0.3.5.
Function exists but segmentation inaccurate	Use option 3.1; new tokenizer in v3 is more stable across languages .
Corporate policy forbids unofficial wheels	Apply option 3.2 (monkey‑patch) and document the loss of filtering.

---

6 Useful references (why some links were not directly used)

• PyPI’s metadata page for 0.2.0 shows no setLangfilters support – used above.
• The PiWheels index confirms only 0.2.0 was built for deb‑based distros .
• Multiple GitHub issue threads document the identical traceback in independent projects GitHub GitHub.
• A Chinese tech Q&A (CSDN) walks through alternative install links for 0.3.5 .

---

Next step for you

Try Option 1 in a fresh Python 3.8/3.9 environment; if the ImportError disappears, your TTS pipeline should resume. Let me know how it goes or if you prefer help migrating to GPT‑SoVITS v3.

Images:

Yes, you can include a GitHub repository directly in your requirements.txt file. To do this, use the following format:

text
git+https://github.com/username/repository.git@branch_or_tag#egg=package_name

For your specific case with LangSegment, you can add:

text
git+https://github.com/chameleon-ai/LangSegment-0.3.5-backup.git@main#egg=LangSegment

Explanation of the components:

• git+https://...: Specifies that pip should install the package from a Git repository over HTTPS.
• @main: Indicates the branch or tag to install from; replace main with the desired branch or tag if different.
• #egg=LangSegment: Names the package; this is necessary for pip to identify the package name correctly.

Additional Tips:

• Ensure that the repository contains a valid setup.py file at its root; pip relies on this to install the package.
• If the package is not published on PyPI, specifying the #egg=package_name is crucial for pip to process the dependency correctly.
• When using this method, pip will clone the repository and install the package as if it were a local project.

By adding this line to your requirements.txt, pip will handle the installation of LangSegment directly from the specified GitHub repository.