AudioGPT
AIGC-Audio/AudioGPTReviewed by AI
Checked by RepoPicks Radar on
What is AudioGPT?
AudioGPT is the code behind a research paper. You write what you want in plain words and a chat model picks which audio model to run: reading text aloud, changing a voice's style, transcribing speech, separating or cleaning up sound, making a sound effect from text or a picture, singing, and a talking head video. Many of the tasks are marked work in progress and the repository stopped in May 2023. It is a Python project for a Linux machine with an NVIDIA card, and the models are downloaded separately.
How easy is it to install AudioGPT?
E · developers only · Linux
The README's whole "Get Started" section is "Please refer to run.md", there are no releases and no installer. requirements.txt pins around 70 packages against the CUDA 11.3 build of PyTorch (torch==1.12.1, mmcv==1.5.0) and pulls one from GitHub, and a download.sh fetches the models, so getting it running is developer work on a Linux machine with an NVIDIA card.
Coming soon: install in 1 click with the RepoPicks app.
AudioGPT on GitHub
- Stars
- 10,167on Nov 10, 2026
- Downloads
- —all releases
- Growth
- +1stars in 6 days
Stars over time
Facts
- Runs on
- Linux
- You need to install
- Python, NVIDIA graphics card, FFmpeg
- To build from source
- Python, NVIDIA graphics card, FFmpeg
- Disk space
- 950 MB–4.3 GB · estimate
- Accounts or keys
- OpenAI
- Cost
- Free, API costs extra · No price for the code, but it is built around a chat model: requirements.txt pins the "openai" and "langchain" packages and the main file is audio-chatgpt.py, so expect to pay for the OpenAI usage.
- License
- NOASSERTION (Other)
- Can replace
- ElevenLabs
- Last change on GitHub
- May 5, 2023
AudioGPT risk signals (4)
- AttentionDependencies with known vulnerabilities
Pinned dependencies with known vulnerabilities (OSV): basicsr: GHSA-86w8-vhw6-q9qq, PYSEC-2026-1215; langchain: GHSA-2qmj-7962-cjq8, GHSA-3644-q5cj-c5c7, GHSA-3hjh-jh2h-vrg6; opencv-contrib-python: GHSA-cxjf-x6jp-p7mc, PYSEC-2023-181; pytorch-lightning: GHSA-2vj5-px25-gjrp, GHSA-4cv3-v7pv-rfhf, GHSA-75m9-98v2-hjpm; soupsieve: GHSA-2wc2-fm75-p42x, GHSA-836r-79rf-4m37, GHSA-gjv8-xp57-g29c; streamlit: GHSA-7p48-42j8-8846, GHSA-8qw9-gf7w-42x5, GHSA-rxff-vr5r-8cj5; torch: GHSA-3749-ghw9-m3mg, GHSA-47fc-vmwq-366v, GHSA-53q9-r3pm-6pq6; transformers: GHSA-282v-666c-3fvg, GHSA-29pf-2h5f-8g72, GHSA-37mw-44qp-f5jm
automatic check
- Attentionno licence granted
The LICENSE file in the repository is empty and GitHub reports no licence for it, while the README says "We provide our implementation and pretrained models as open source". With no licence text, nobody has given you permission to use or redistribute the code.
AI review
- Infounfinished and left alone
The capability tables mark many tasks "Yes (WIP)" or "WIP", the README itself says "Currently not every model has repository", and the last commit is from 2023-05-05. Expect missing pieces and no fixes.
AI review
- Infopulls code and models from elsewhere
requirements.txt starts with "--extra-index-url https://download.pytorch.org/whl/cu113" and ends with "git+https://github.com/openai/CLIP.git", and the repository has a download.sh for the model files. The install reaches out to several outside places.
AI review
Evidence is shown in its original language.
What Radar could not check about AudioGPT
- Whether the install still works at all, since the pinned versions are from 2023
- Whether the Hugging Face demo page listed as the project homepage still runs
- How much disk space the downloaded models take
- Which tasks actually work, given how many are marked work in progress
- How much the OpenAI usage costs for a session