vinayapathak commited on
Commit
af7d6fc
路
verified 路
1 Parent(s): 23fa293

Release notarized Spum CLI 0.2.3 with native offline speech

Browse files
Files changed (4) hide show
  1. LICENSES.md +11 -0
  2. README.md +30 -24
  3. SHA256SUMS.txt +1 -0
  4. spum-0.2.3-macos-arm64.tar.gz +3 -0
LICENSES.md ADDED
@@ -0,0 +1,11 @@
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # Licences and attribution
2
+
3
+ This repository distributes multiple artifacts with different licences. The repository-level LICENSE applies to Spum software; it does not relicense the bundled third-party components or model weights.
4
+
5
+ - Spum software: MIT; see LICENSE.
6
+ - Whisper base.en model weights: MIT, copyright OpenAI. The corresponding model archive includes the original notice.
7
+ - Kokoro 1.0 model and voice embeddings: Apache-2.0, as published by hexgrad. The model archive includes its licence and identifies the upstream ONNX conversion.
8
+ - Misaki pronunciation data: Apache-2.0. Supplementary CMUdict pronunciation data: BSD-style with the original Carnegie Mellon notice. Both notices and the source hashes are inside the synthesis archive.
9
+ - Native speech runtime: sherpa-onnx (Apache-2.0), ONNX Runtime (MIT), and their attributed dependencies. The CLI archive includes THIRD_PARTY_NOTICES.md, original licence files, and unmodified public source for the included covered components. The Eigen build uses EIGEN_MPL2_ONLY. No proprietary Spum source is included in the third-party source archive.
10
+
11
+ No espeak runtime or espeak pronunciation data is included in these speech packages. No private patient images, restricted training datasets or private adapted model weights are distributed here. See each archive's PROVENANCE.json and LICENSES directory for details. Retain the applicable notices when redistributing an artifact.
README.md CHANGED
@@ -1,53 +1,59 @@
1
  ---
2
- license: mit
 
 
3
  language:
4
  - en
5
  tags:
6
  - local-inference
7
  - apple-silicon
8
  - biomedical
9
- - vision-language
 
 
10
  ---
11
  # Spum
12
 
13
- **Physical AI Engine for Biomedical Inference**
14
 
15
- Scientific Perception Uniting Mind & Matter. A Neuroscape Imaging product.
16
 
17
- This repository distributes compiled Spum releases. It does not contain model weights or training data. Download compatible GGUF weights separately under their own licenses.
18
 
19
- ## Mac app
20
 
21
- Download `Spum-0.2.2-macOS-arm64.dmg`, open it and drag Spum to Applications. The native app includes local chat, image attachments, a model library, conversation search and export, light and dark themes, and a menu-bar shortcut. Requires Apple silicon and macOS 14 or later. The app is Developer ID signed and notarized by Apple.
 
 
22
 
23
  ```sh
24
- brew install --cask vinayapathak/spum/spum-app
25
- ```
26
-
27
- ## Command line
28
-
29
- ```sh
30
- curl -fsSL https://nsimaging.ai/spum/install.sh | sh
31
  # or
32
  brew install vinayapathak/spum/spum
33
  ```
34
 
35
- The archive `spum-0.2.1-macos-arm64.tar.gz` contains the signed command line engine and its license. The Mac app version is 0.2.2; its bundled inference engine is 0.2.1.
36
 
37
- ## Run an image model
38
 
39
- Supply a compatible model and its matching image projector.
 
 
 
40
 
41
  ```sh
42
- spum run medgemma-4b-it-Q4_K_M.gguf --gpu --image chest-xray.jpg -p "Describe this image in one sentence."
 
 
 
 
 
43
  ```
44
 
45
- Outputs depend on the selected model. Spum is a research and engineering tool, not a medical device; these releases do not establish diagnostic or segmentation accuracy.
46
 
47
- ## Links
48
 
49
- - [Website and recorded runs](https://nsimaging.ai/spum/)
50
- - [Mac installation](https://nsimaging.ai/spum/#install)
51
- - [Homebrew tap](https://github.com/vinayapathak/homebrew-spum)
52
 
53
- Founder: Vinayak Pathak, Biomedical Engineer, Duke University.
 
1
  ---
2
+ license: other
3
+ license_name: spum-artifact-licenses
4
+ license_link: https://huggingface.co/vinayapathak/spum/blob/main/LICENSES.md
5
  language:
6
  - en
7
  tags:
8
  - local-inference
9
  - apple-silicon
10
  - biomedical
11
+ - automatic-speech-recognition
12
+ - text-to-speech
13
+ - onnx
14
  ---
15
  # Spum
16
 
17
+ **A Neuroscape Imaging product.** Published by Neuroscape Imaging Private Limited.
18
 
19
+ Spum brings local inference to the computer beside the instrument. This repository contains signed Mac software releases and tested, openly licensed speech model packages. It does not contain proprietary engine source or private medical data.
20
 
21
+ Website: [spum.si](https://spum.si/) 路 [Documentation](https://spum.si/spum/docs/) 路 [Speech guide](https://spum.si/spum/docs/speech/) 路 [Whitepaper](https://spum.si/spum/whitepaper/spum-whitepaper.pdf)
22
 
23
+ ## Software releases
24
 
25
+ - [CLI 0.2.3 for Apple silicon](spum-0.2.3-macos-arm64.tar.gz): includes the signed native speech runtime, notarized by Apple.
26
+ - [Standalone Mac app 0.2.2](Spum-0.2.2-macOS-arm64.dmg): a separately versioned native app. Speech in CLI 0.2.3 is not a new desktop-app feature.
27
+ - Archive hashes: [SHA256SUMS.txt](SHA256SUMS.txt).
28
 
29
  ```sh
30
+ curl -fsSL https://spum.si/install.sh | sh
 
 
 
 
 
 
31
  # or
32
  brew install vinayapathak/spum/spum
33
  ```
34
 
35
+ Keep the included native runtime with the CLI, or use the installer/Homebrew package. No Python runtime is required.
36
 
37
+ ## Released speech models
38
 
39
+ | ID | Task | Upstream |
40
+ |---|---|---|
41
+ | `whisper-base-en` | English speech recognition | OpenAI Whisper base.en, int8 ONNX conversion |
42
+ | `kokoro-en-v1-int8` | English speech synthesis | Kokoro 1.0 int8, three checked English voices |
43
 
44
  ```sh
45
+ spum models catalog
46
+ spum models download whisper-base-en
47
+ spum models download kokoro-en-v1-int8
48
+ spum models verify whisper-base-en
49
+ spum transcribe recording.wav --json --output transcript.json
50
+ spum speak --text "The scanner is ready." --voice af_heart --output speech.wav
51
  ```
52
 
53
+ See [models/](models) for model archives, their catalog, provenance, original licences and the small functional-check record. These are third-party pretrained models packaged for Spum; Neuroscape does not claim to have trained the upstream foundation models. Model files have separate licences described in [LICENSES.md](LICENSES.md), not a blanket repository licence.
54
 
55
+ The recognition pack is English only. The synthesis interface exposes `af_heart`, `af_alloy` and `am_adam`, produces 24 kHz mono WAV, and checks pronunciation coverage before generating audio. Captions use recognition-window timing rather than word alignment. No microphone, remote speech API or voice-cloning facility is used.
56
 
57
+ ## Evaluation scope
 
 
58
 
59
+ Three reference recordings contained 77 normalized words with 3 word differences. Three synthesized phrases produced finite, non-silent audio, and an ASR roundtrip recovered their words. These small functional checks are not a general accuracy benchmark or an independent perceptual listening study. Spum is a research and engineering tool; these demonstrations do not establish clinical transcription or diagnostic validity.
SHA256SUMS.txt CHANGED
@@ -1,2 +1,3 @@
 
1
  dad3f3fa1a6af4a83db5e3317b934cedcf708f41c0d22a094e15768f94c7ea6f Spum-0.2.2-macOS-arm64.dmg
2
  ac684d647823449b49d82b07e113f191000158d166b2c1e59e1bef8d07fba36d spum-0.2.1-macos-arm64.tar.gz
 
1
+ ebf325321e5e2813307cd75781fad3b6f52f79a96fd5d6aace7a30734c132682 spum-0.2.3-macos-arm64.tar.gz
2
  dad3f3fa1a6af4a83db5e3317b934cedcf708f41c0d22a094e15768f94c7ea6f Spum-0.2.2-macOS-arm64.dmg
3
  ac684d647823449b49d82b07e113f191000158d166b2c1e59e1bef8d07fba36d spum-0.2.1-macos-arm64.tar.gz
spum-0.2.3-macos-arm64.tar.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ebf325321e5e2813307cd75781fad3b6f52f79a96fd5d6aace7a30734c132682
3
+ size 13118930