VibeVoice

Open-Source Frontier Voice AI

Runssource: GitHubhomepagePythonMITcommit 16fb2cb1217c

Python, MIT licensed.

VibeVoice runs. An Argusic agent installed it in 2 minutes and hit 5 errors on a clean machine with no GPU, and the whole session was recorded.

The editable install succeeds, all 86 repository tests pass, the realtime TTS demo launches and serves HTTP 200/101, and real model inference produced a decoded non-silent audio file; real ASR inference could not load within the 8 GB memory limit.

What the agent ended up with on a clean machine, in its own words. How this is measured

Measured by Argusic on a fresh machine every time. Every number links to its evidence. Argusic Score 94 of 100 (the mean of this project's run scores).

At a glance

verdict
Runs
Argusic Score
94 / 100
cost of the verifying run
$0.43 (measured)
recorded runs
2
last tested
stars
56,956
forks
6,416
open issues
205
watchers
273
size
211 MB
created
last push

Subject data from GitHub, linked at the top of this page, refreshed . Test data by Argusic (CC BY 4.0); every number links to a run page with the full log, the recording, and their sha256 hashes.

What was measured

From the run that decided this verdict. Every line is a recorded value, not an estimate.

Time lost to each failure while testing VibeVoice, drawn from the recorded run
Every bar is a failure the agent recorded, with the time it cost. Drawn from this run, not from anywhere else.

What went wrong, in the agent's own words

The agent's own words, worst first, never rewritten into advice.

  • Bundled voice presets failed safe torch.load because BaseModelOutputWithPast and the legacy DynamicCache layout were incompatible with PyTorch 2.6+/Transformers 4.57.9 minutes
  • Real 7B ASR checkpoint load was OOM-killed in float32 and BFloat16 because the container cgroup is limited to 8 GB and the checkpoint is 15.5 GB.6 minutes
  • Real ASR checkpoint tokenizer rejected its list-valued extra_special_tokens under Transformers 4.57.3 minutes
  • `python` and the `time` shell utility were absent; creating a virtualenv at /work/venv failed with permission denied.1 minute
  • ASR CLI --help crashed because torch.backends.xpu did not exist in the installed PyTorch.1 minute
Install time
2 minutes
Cold machine to finish
27 minutes
Errors hit and fixed
5 hit, 4 fixed with no human help
How the result was proved
`/tmp/vvvenv/bin/python -m pytest -q tests` exited 0 with `86 passed`; `ffmpeg -v error -i /tmp/vvout3/short_generated.wav -f null -` exited 0 after `realtime_model_inference_from_file.py` exited 0 and saved 1.733333 seconds of audio; `curl` to localhost:3000 returned 200 for `/` and `/config`, and 101 for `/stream`.
Model tokens used
529,780
Exact commit tested
16fb2cb1217c
Container image
sha256:cdd920bce783

What was not measured

The other half, so you can see exactly how far this evidence goes. A gap is printed as a gap and never filled in.

Peak GPU memory
this ran on a CPU machine, so there was no GPU to sample
Behaviour over a long run
each attempt uses the smallest documented settings and a fixed time budget, so this measures whether it runs, not how it holds up under load

Watch the session

The whole run, from a clean machine, at the speed it actually happened. Nothing is cut and nothing is sped up.

A recording of attempt 1, run . The verdict above comes from a different attempt, so read this as one session and not as the current result. The terminal recording it was made from is the evidence, stored with a sha256 fingerprint; the video is a rendering of it. Nothing loads from the video host until you press play.

The end of the recording

The last lines of the terminal session that decided this verdict, exactly as they were recorded.

 .../modeling_vibevoice_streaming_inference.py      | 42 ++++++++++++++++++++++
 .../modular/modular_vibevoice_text_tokenizer.py    | 10 ++++++
 5 files changed, 61 insertions(+), 14 deletions(-)
 M demo/realtime_model_inference_from_file.py
 M demo/vibevoice_asr_inference_from_file.py
 M demo/web/app.py
 M vibevoice/modular/modeling_vibevoice_streaming_inference.py
 M vibevoice/modular/modular_vibevoice_text_tokenizer.py
The repository now installs, its tests pass, and the documented realtime TTS path generates audio t…
tokens used
529,780
The repository now installs, its tests pass, and the documented realtime TTS path generates audio t…

Replay the whole session, every command from a clean machine to this point.

How it was tested

One agent, one container, no human help. It installs the project the way its own documentation says to, runs it, and fixes what breaks. Everything below is recorded as it happened: the terminal session, the log and the exact commit. The full method.

Strengths and limits

Measured facts, not opinions. How this is written.

What went well

  • Reached a running state on a clean machine, with the session recorded.
  • Installed in 2 minutes, faster than the median of the 46 comparable projects Argusic has measured.
  • Ran without a GPU, so it does not need one to start.
  • Licensed MIT, as reported by its host.
  • Measured 2 times, so the result is not a one-off.

What did not, or is not known

  • Hit 5 errors during setup, 1 of which were not solved.

Of the 47 Python projects Argusic has installed and timed, VibeVoice was the 21st fastest to reach a running state, and 31 of 47 reached one at all.

What it is a reasonable choice for

  • Trying it on a laptop or a small server: it reached a running state without a GPU.
  • A quick evaluation: it was running 2 minutes after the clone.

Run history

tested byattemptoutcomescorecost USDwhen
Argusic Runner2/3Runs96.000.43
Argusic Runner1/3Runs with mocks92.000.19

Embed the badge

Markdown for the project README. It links back here; terms on the terms page.

[![Tested by Argusic](https://argusic.com/badge/VibeVoice.svg)](https://argusic.com/subject/vibevoice)

Questions

Does VibeVoice run?
Yes. VibeVoice runs. Argusic installed and launched it on a clean machine in 2 minutes, hitting 5 errors on the way, and recorded the session.
How did Argusic test VibeVoice?
On a fresh, disposable machine, with every command recorded and the repository pinned at commit 16fb2cb1217c. 2 attempts are recorded, and the full method is on the methodology page.
What did it cost to test VibeVoice?
The run that produced this verdict cost $0.43: real compute and model cost, not a list price per million tokens. It is recorded on that run's page.
How long does VibeVoice take to install?
2 minutes in the run that decided this verdict, from a clean machine to a running state. That is one measured attempt, not an average.
Does VibeVoice need a GPU?
Not to start. Argusic reached a running state on a machine with no GPU.
What goes wrong when installing VibeVoice?
5 things broke in the recorded run, and 4 were fixed without human help. Each one, and the time it cost, is listed on this page.
How does VibeVoice compare with the alternatives?
Of the 47 Python projects Argusic has installed and timed, VibeVoice was the 21st fastest to reach a running state, and 31 of 47 reached one at all.
Where is the evidence for VibeVoice?
All 2 recorded runs are on this page, each linking to its full log and terminal recording, stored with a sha256 fingerprint so it cannot be quietly altered.

Discussion