mirror of
https://github.com/anomalyco/hex.git
synced 2026-11-22 21:23:02 +01:00
* add fluid speech pro to the speech model catalog FluidVoice's Fluid Speech Pro, a Parakeet TDT 0.6B v3 retrain that also declares Moondream's Parakeet Ultra as a base, for the same 25 European languages. TalkMate converted it to a Q8_0 transcribe.cpp GGUF at d59dc79 (741,819,456 bytes, sha256 c3848e98...); the file says parakeet/tdt-0.6b-fluid-speech-pro and announces its extra pre-Conformer block as stt.parakeet.encoder.pre_block_kernel, so it needs a transcribe.cpp that reads enc.pre_block. Like Parakeet v3 it takes no language hint, detects the language under Auto, does not stream, and is never recommended or badged. Its card publishes no LibriSpeech test-clean figure, so it carries no single word error rate and no benchmark record; its per-language FLEURS figures and the CC BY 4.0 credit are in its details. * bump(transcribe.cpp): 295c7e6, fluid speech pro's pre-conformer block Fluid Speech Pro's GGUF announces stt.parakeet.encoder.pre_block_kernel and carries eight enc.pre_block.* tensors: a residual FrameCleaner block between the subsampler and Conformer layer 0. The old pin's graph never read them, so the model loaded as plain Parakeet v3 and its encoder was wrong. Both transcribe-cpp crates move to 295c7e601 on hex/combined-fluid, a fast-forward of bcb639f that runs the block on pre_encode before xscaling when the KV is present, and fails the load rather than skipping tensors when it is only partly there. Cargo.lock changes only those two sources (`getrandom`'s tempfile pointer was reverted to what it was), and nix/package.nix carries the hash from a failed `nix build .#hex`, then a full one whose check phase passed. `hex model install fluid-speech-pro` followed by `hex model transcribe` on whisper.cpp's jfk.wav prints the expected sentence, as does Parakeet v3 on the same clip. * fluid speech pro: snake_case id, FLEURS-only figures, blurb, notices
226 lines
12 KiB
Markdown
226 lines
12 KiB
Markdown
# Third-Party Notices
|
||
|
||
Hex's own source is covered by the root `LICENSE`. The following notices cover
|
||
vendored icon assets and the speech models Hex downloads under their own
|
||
licences. Dependencies retain their respective upstream licenses.
|
||
|
||
## Parakeet Ultra
|
||
|
||
Parakeet Ultra by Moondream, based on NVIDIA Parakeet TDT 0.6B v3. GGUF
|
||
conversion and quantization by handy-computer. Creative Commons Attribution
|
||
4.0 International (CC BY 4.0). Hex downloads `parakeet-ultra-Q8_0.gguf` from
|
||
https://huggingface.co/handy-computer/parakeet-ultra-gguf/tree/a1539be1b849ea7463b88343e5277b373d0c4ea9
|
||
only while Labs › Parakeet Ultra is on; it does not ship the weights.
|
||
|
||
License: https://creativecommons.org/licenses/by/4.0/legalcode.en
|
||
|
||
Notice supplied with the files:
|
||
|
||
> Modifications: handy-computer converted
|
||
> https://huggingface.co/moondream/parakeet-ultra at commit 73175eb to GGUF for
|
||
> transcribe.cpp (https://github.com/handy-computer/transcribe.cpp). Conversion
|
||
> repackages the weights, renaming and re-laying-out tensors; the quantized
|
||
> files (Q8_0, Q6_K, Q5_K_M, Q4_K_M) also store them at reduced numeric
|
||
> precision. The weights were not retrained or fine-tuned.
|
||
|
||
The weights are provided as-is, without warranties, as the licence states.
|
||
|
||
## Orukeet
|
||
|
||
Orukeet by Oruk AI, based on NVIDIA Parakeet TDT 0.6B v3. GGUF conversion and
|
||
quantization by handy-computer. Creative Commons Attribution-ShareAlike 4.0
|
||
International (CC BY-SA 4.0). Hex downloads `orukeet-Q8_0.gguf` from
|
||
https://huggingface.co/handy-computer/orukeet-gguf/tree/c5511a3850fce1cbbfa5d94ef815966e1d99275f
|
||
and does not ship the weights.
|
||
|
||
License: https://creativecommons.org/licenses/by-sa/4.0/legalcode.en
|
||
|
||
Notice supplied with the files:
|
||
|
||
> # Attribution and license scope
|
||
>
|
||
> Orukeet is an adaptation of **NVIDIA Parakeet TDT 0.6B v3**. NVIDIA retains
|
||
> copyright in its model and upstream work. The base weights are distributed
|
||
> under [CC BY 4.0](https://huggingface.co/nvidia/parakeet-tdt-0.6b-v3).
|
||
> Oruk AI's changes comprise multilingual/accent continuation training,
|
||
> parameter blending, fitted and frozen Gabor-kernel replacement and recovery,
|
||
> native and ONNX export, application integration and evaluation.
|
||
> [Model stages](https://github.com/Oruk-AI/orukeet/blob/main/release/model-stages.json)
|
||
> identifies r3 as the source of every current Orukeet download and records its
|
||
> ancestry.
|
||
>
|
||
> The r3 NeMo, ONNX INT8, Q8 GGUF and F16 GGUF weights, and their fitted Gabor
|
||
> kernels, are designated **CC BY-SA 4.0**. The complete license is in
|
||
> LICENSE-WEIGHTS. This permits commercial use and modification, with
|
||
> attribution and applicable ShareAlike requirements.
|
||
> Earlier checkpoints retain their source notices and are not silently
|
||
> relicensed by this file. Orukeet v0.1.0 distributes the final r3 checkpoint in
|
||
> all four formats.
|
||
>
|
||
> Python and integration code in this repository is MIT unless a file specifies
|
||
> otherwise. The native bindings, audio windowing and worker transport derive
|
||
> from OpenWhisper by Knuckles92 and its Oruk AI integration; their MIT notice
|
||
> is retained in
|
||
> [LICENSE](https://huggingface.co/handy-computer/orukeet-gguf/blob/c5511a3850fce1cbbfa5d94ef815966e1d99275f/LICENSE).
|
||
> The downloaded NVIDIA NeMo-Speech.cpp SDK
|
||
> and its bundled dependencies retain their own notices, including Apache-2.0
|
||
> and MIT components. Keep the SDK's license files when redistributing it. The
|
||
> [pinned source](https://github.com/NVIDIA/NeMo-Speech.cpp/tree/4f9676226f667d14608487df744f375db87127f8)
|
||
> is the authority for those terms.
|
||
>
|
||
> The ONNX exporter follows sherpa-onnx's Parakeet TDT v3 conversion script. Its
|
||
> upstream reference and Apache-2.0 license are retained in
|
||
> [export/onnx](https://github.com/Oruk-AI/orukeet/blob/main/export/onnx/README.md).
|
||
> The ONNX archive includes the weight license and source attribution. The
|
||
> optimized ONNX encoder uses exactly equivalent FP32 arithmetic for 24
|
||
> quantized depthwise convolutions, with existing runtime operators and
|
||
> unchanged quantized values. Original export and separate optimization
|
||
> receipts are retained in the GitHub repository under
|
||
> `evidence/onnx-r3-20260910/` and `evidence/speed20260910/`.
|
||
>
|
||
> Training data credits:
|
||
>
|
||
> - Mozilla/Common Voice contributors: Common Voice 22, CC0.
|
||
> - Google and the FLEURS authors: FLEURS, CC BY 4.0.
|
||
> - Google and the OpenSLR 83 authors, with the `ylacombe/english_dialects`
|
||
> restructuring: English dialect speech, CC BY-SA 4.0.
|
||
> - Beijing Kingline Data Technology and the SpeechOcean762 authors:
|
||
> SpeechOcean762, CC BY 4.0.
|
||
> - SpeechColab and the GigaSpeechBench authors: 17 English accent/domain splits
|
||
> in the final continuation. Their paper identifies Creative Commons source
|
||
> audio; the source record below preserves the available license details.
|
||
> - DISCO at ETH Zurich and the EuroSpeech contributors: Bulgarian, Greek and
|
||
> Italian parliamentary speech, with the providers' country-specific terms.
|
||
> - SberDevices and the Golos authors: Golos Crowd, under the
|
||
> [Public license with attribution and conditions reserved](https://github.com/sberdevices/golos/blob/master/license/en_us.pdf).
|
||
> - Nordisk Språkteknologi, the National Library of Norway and the Alexandra
|
||
> Institute: NST Swedish and Danish, distributed under CC0.
|
||
> - ILSP/Athena Research Center and the Lesbian Speech Corpus contributors:
|
||
> dialect speech from Lesbos. The source card does not specify a reuse
|
||
> license.
|
||
>
|
||
> Data were filtered, split, normalized and sampled as described in
|
||
> [data and licenses](https://github.com/Oruk-AI/orukeet/blob/main/docs/data-and-licenses.md).
|
||
> Corpus copyrights and source terms remain with their owners. Evaluation-only
|
||
> data have separate terms. No endorsement by NVIDIA, Mozilla, Google or the
|
||
> other source projects is implied.
|
||
>
|
||
> The generated statistical records in `evidence/metric-evidence.tar.gz`, the
|
||
> Gabor recovery count bundles under `training/gabor_half/results/`, and the
|
||
> FT-4035 benchmark counts under `evidence/regression-ft-20260907/` are
|
||
> designated CC BY 4.0, attributed to Oruk AI. This designation covers the
|
||
> prepared metric records; it does not relicense the original corpus
|
||
> recordings or transcripts, which are not included in that archive.
|
||
>
|
||
> The r3 numeric benchmark records in `evidence/standard-asr-r3-20260908/` and
|
||
> `evidence/domains-r3-20260908/` are released under CC BY 4.0. They contain
|
||
> error counts and recording identifiers; dataset audio remains with its
|
||
> providers.
|
||
>
|
||
> The paired ONNX application benchmark counts in
|
||
> `evidence/onnx-r3-20260910/` are also released under CC BY 4.0. These records
|
||
> contain numeric errors, timings and recording identifiers, without corpus
|
||
> audio or transcripts.
|
||
>
|
||
> Modifications: handy-computer converted https://huggingface.co/oruk/orukeet
|
||
> at commit b59c13a to GGUF for transcribe.cpp
|
||
> (https://github.com/handy-computer/transcribe.cpp). Conversion repackages the
|
||
> weights, renaming and re-laying-out tensors; the quantized files (Q8_0,
|
||
> Q6_K, Q5_K_M, Q4_K_M) also store them at reduced numeric precision. The weights
|
||
> were not retrained or fine-tuned.
|
||
|
||
The weights are provided as-is, without warranties, as the licence states.
|
||
|
||
## Fluid Speech Pro
|
||
|
||
Fluid Speech Pro by FluidVoice, based on NVIDIA Parakeet TDT 0.6B v3 and
|
||
Moondream Parakeet Ultra. GGUF conversion and quantization by TalkMate.
|
||
Creative Commons Attribution 4.0 International (CC BY 4.0). Hex downloads
|
||
`fluid-speech-pro-Q8_0.gguf` from
|
||
https://huggingface.co/TalkMate/fluid-speech-pro-gguf/tree/d59dc79f7e04715c5b0ba07f01106f71ebea97dc
|
||
and does not ship the weights.
|
||
|
||
Copyright (c) 2026 FluidVoice. Licensed under CC BY 4.0 (see LICENSE).
|
||
|
||
Attribution for use: "Fluid Speech Pro by FluidVoice. Built on NVIDIA Parakeet TDT 0.6B v3 and Moondream Parakeet Ultra."
|
||
|
||
License: https://creativecommons.org/licenses/by/4.0/legalcode.en
|
||
|
||
Notice supplied with the files:
|
||
|
||
> Fluid Speech Pro by FluidVoice
|
||
> (https://huggingface.co/FluidVoice-Labs/fluid-speech-pro), CC BY 4.0.
|
||
> Built on NVIDIA Parakeet TDT 0.6B v3 and Moondream Parakeet Ultra, both CC
|
||
> BY 4.0. Converted to a transcribe.cpp GGUF (Q8_0) by TalkMate, in the layout
|
||
> of Handy's Parakeet Ultra GGUF
|
||
> (https://huggingface.co/handy-computer/parakeet-ultra-gguf, CC BY 4.0).
|
||
|
||
The weights are provided as-is, without warranties, as the licence states.
|
||
|
||
## Wayland Input Method Protocol
|
||
|
||
`src/linux/protocols/input-method-unstable-v2.xml` is the upstream
|
||
`input-method-unstable-v2` protocol description, copied unchanged from
|
||
wayland-protocols-misc 0.3.12. It is MIT licensed; the copyright header is
|
||
kept in the file.
|
||
|
||
Upstream: https://gitlab.freedesktop.org/wayland/wayland-protocols
|
||
|
||
## Lucide And Feather
|
||
|
||
The settings sidebar icons in `assets/icons/` (`settings`, `modes`,
|
||
`meetings`, `history`, `voice-filter`, `scrawl`, `hud-lab`, `activity`) are
|
||
Lucide 1.62.0's `sliders-horizontal`, `layers`, `users`, `history`,
|
||
`audio-waveform`, `scan-line`, `flask-conical`, and `activity`, with stroke
|
||
attributes moved onto each element and a 1.75 stroke.
|
||
The `assets/icons/lucide/*.svg` assets in the Windows development branch are
|
||
also from Lucide: https://github.com/lucide-icons/lucide
|
||
|
||
Upstream license: https://github.com/lucide-icons/lucide/blob/23f9abc4ed0146cffededd3d7f94c1018bfdf693/LICENSE
|
||
|
||
`search.svg` has a modified handle endpoint and reordered SVG elements. The
|
||
complete upstream license below covers both original and derived assets.
|
||
|
||
ISC License
|
||
|
||
Copyright (c) 2026 Lucide Icons and Contributors
|
||
|
||
Permission to use, copy, modify, and/or distribute this software for any
|
||
purpose with or without fee is hereby granted, provided that the above
|
||
copyright notice and this permission notice appear in all copies.
|
||
|
||
THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
|
||
WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
|
||
MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
|
||
ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
|
||
WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
|
||
ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
|
||
OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
|
||
|
||
---
|
||
|
||
The following Lucide icons are derived from the Feather project:
|
||
|
||
airplay, alert-circle, alert-octagon, alert-triangle, aperture, arrow-down-circle, arrow-down-left, arrow-down-right, arrow-down, arrow-left-circle, arrow-left, arrow-right-circle, arrow-right, arrow-up-circle, arrow-up-left, arrow-up-right, arrow-up, at-sign, calendar, cast, check, chevron-down, chevron-left, chevron-right, chevron-up, chevrons-down, chevrons-left, chevrons-right, chevrons-up, circle, clipboard, clock, code, columns, command, compass, corner-down-left, corner-down-right, corner-left-down, corner-left-up, corner-right-down, corner-right-up, corner-up-left, corner-up-right, crosshair, database, divide-circle, divide-square, dollar-sign, download, external-link, feather, frown, hash, headphones, help-circle, info, italic, key, layout, life-buoy, link-2, link, loader, lock, log-in, log-out, maximize, meh, minimize, minimize-2, minus-circle, minus-square, minus, monitor, moon, more-horizontal, more-vertical, move, music, navigation-2, navigation, octagon, pause-circle, percent, plus-circle, plus-square, plus, power, radio, rss, search, server, share, shopping-bag, sidebar, smartphone, smile, square, table-2, tablet, target, terminal, trash-2, trash, triangle, tv, type, upload, x-circle, x-octagon, x-square, x, zoom-in, zoom-out
|
||
|
||
The MIT License (MIT) (for the icons listed above)
|
||
|
||
Copyright (c) 2013-present Cole Bemis
|
||
|
||
Permission is hereby granted, free of charge, to any person obtaining a copy
|
||
of this software and associated documentation files (the "Software"), to deal
|
||
in the Software without restriction, including without limitation the rights
|
||
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
|
||
copies of the Software, and to permit persons to whom the Software is
|
||
furnished to do so, subject to the following conditions:
|
||
|
||
The above copyright notice and this permission notice shall be included in all
|
||
copies or substantial portions of the Software.
|
||
|
||
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
||
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
||
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
|
||
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
||
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
||
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
|
||
SOFTWARE.
|