TheWhisper

Optimized Whisper models for streaming and on-device use

Runs with mockssource: GitHubhomepagePythonMITcommit af7eebaa5315

Python, MIT licensed. The project labels itself: apple silicon, coreml, mlx, nvidia gpu, on device ai, real time, speech recognition and speech to text.

TheWhisper runs, with stand-ins for the services it depends on. An Argusic agent installed it in 35 minutes and hit 7 errors on a clean machine with no GPU, and the whole session was recorded.

In this container the Python package installs and imports, the streaming pipeline and examples/server.py run against a local mock ASR endpoint returning word timestamps, and the Electron client launches on the virtual display and completes a session-create/audio-chunk/process transcription round trip; real CUDA/CoreML model inference and microphone capture were not verifiable here.

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 86.3 of 100 (the mean of this project's run scores).

At a glance

verdict
Runs with mocks
Argusic Score
86.3 / 100
cost of the verifying run
$0.80 (measured)
recorded runs
1
last tested
stars
904
forks
58
open issues
0
watchers
12
size
3 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 TheWhisper, 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.

  • VAD streaming path silently dropped pending audio, and oversized queued chunks grew the buffer without trimming20 minutes
  • Electron .env loaded from repo root but README runs from electron_app dir8 minutes
  • OSError: PortAudio library not found (system has no libportaudio.so, no root to install)5 minutes
  • Electron getUserMedia audio init returned {} in container (no audio device)4 minutes
  • ModuleNotFoundError: transformers on import thestage_speechkit3 minutes
  • FileStream used librosa correctly but had typo orgi_sr and imported load unnecessarily3 minutes
Install time
35 minutes
Cold machine to finish
34 minutes
Errors hit and fixed
7 hit, 5 fixed with no human help
How the result was proved
python StreamingPipeline remote-backend test printed 7 transcription-bearing results; curl POST localhost:8822/session/create/ returned HTTP 200 and session_id; 40x add_chunk + process returned uncommited_words [{'text':' mock transcription','start':0.0,'end':1.0}]; electron --no-sandbox . on DISPLAY=:99 stayed alive with window 'TheNotes App' and its own api.createSession/sendAudioChunk/process returned the transcription
Model tokens used
3,825,113
Exact commit tested
af7eebaa5315
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
Behaviour against real services
this run was verified against stand-in services only

The end of the recording

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

- Electron audio/mic init unreachable here (no audio device); Apple/`mlx` backend unavailable on Li…
**Changes left in the repo** (uncommitted): `pyproject.toml` (add `transformers`), `thestage_speech…
tokens used
3,825,113
All required work is done and recorded. Final summary with the required line:
**Verified work (all observed in this container, 2026-10-12):**
- `pip install .` in a fresh venv built the wheel and installed cleanly after fixes.
- `StreamingPipeline` remote backend: 5 s audio in 0.1 s steps → 7 transcription-bearing results wi…
- `examples/server.py` (whisper backend vs local mock): `POST /session/create/` → 200, 40 `add_chun…
- Electron app on Xvfb: window `TheNotes App` (1440x900) stayed up; session created via its own `ap…
- Electron audio/mic init unreachable here (no audio device); Apple/`mlx` backend unavailable on Li…
**Changes left in the repo** (uncommitted): `pyproject.toml` (add `transformers`), `thestage_speech…

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

  • Licensed MIT, as reported by its host.

What did not, or is not known

  • Only reached a running state against stand-in services, so nothing here proves it works against the real ones.
  • Took 35 minutes to install, slower than the median of the 7 comparable projects Argusic has measured.
  • Hit 7 errors during setup, 2 of which were not solved.
  • Measured once so far. A single attempt shows it can run, not that it runs reliably.

Of the 8 apple-silicon projects Argusic has installed and timed, TheWhisper was the 8th fastest to reach a running state, and 2 of 8 reached one at all.

Also tested, in the same area

Every one of these was installed and run by Argusic on a clean machine. Nothing appears here that was not tested.

Run history

tested byattemptoutcomescorecost USDwhen
Argusic Runner1/3Runs with mocks86.290.80

Topics (from GitHub)

apple-siliconcoremlmlxnvidia-gpuon-device-aireal-timespeech-recognitionspeech-to-textstreamingtranscriptiontranslationvoicevoice-ai

Embed the badge

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

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

Questions

Does TheWhisper run?
TheWhisper runs, with mocks standing in for external services. Argusic installed and launched it on a clean machine in 35 minutes, hitting 7 errors on the way, and recorded the session.
How did Argusic test TheWhisper?
On a fresh, disposable machine, with every command recorded and the repository pinned at commit af7eebaa5315. One attempt is recorded, and the full method is on the methodology page.
What did it cost to test TheWhisper?
The run that produced this verdict cost $0.80: real compute and model cost, not a list price per million tokens. It is recorded on that run's page.
How long does TheWhisper take to install?
35 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 TheWhisper need a GPU?
Not to start. Argusic reached a running state on a machine with no GPU.
What goes wrong when installing TheWhisper?
7 things broke in the recorded run, and 5 were fixed without human help. Each one, and the time it cost, is listed on this page.
How does TheWhisper compare with the alternatives?
Of the 8 apple-silicon projects Argusic has installed and timed, TheWhisper was the 8th fastest to reach a running state, and 2 of 8 reached one at all.
Where is the evidence for TheWhisper?
The recorded run is on this page, each linking to its full log and terminal recording, stored with a sha256 fingerprint so it cannot be quietly altered.

Discussion