Clear
Studio sound, no cloud bill.
On-device speech enhancement: denoise, dereverb, and loudness-normalize.
- SDKs, install and examples: https://github.com/Desert-Ant-Labs/desert-ant-core/blob/main/docs/models/clear.md
- Website: https://desertant.com/models/clear/
Record in a coffee shop and get studio sound, for podcasters, video creators and voice apps.
Reduce noise, clean up audio, normalize volume. Works on noisy 48kHz mono or stereo audio: meeting recorders, bluetooth microphones, built-in phone mics, a laptop in a coffee shop.
Runs entirely on the Apple Neural Engine via Core ML and on Android via ONNX Runtime. The Core ML assets use the iOS 16 model format; the Swift SDK currently supports iOS 17+ and macOS 14+.
Sound
Trained to deliver a rich, present, close-miked podcast sound.
- Denoised. HVAC, keyboard clicks, mouse rustle, mic bumps, room hum, laptop fans, coffee shop background, all pulled down without chewing consonants.
- Dereverbed. Untreated bedrooms, offices and hotel rooms come out sounding closer to a treated studio. The model does not add reverberation of its own.
- Warm and present. Low-mids brought forward so voice sits comfortably in a mix rather than sounding thin or distant.
- Sibilance-safe. No harsh peaks introduced when cleaning up S / T / F consonants.
- No pumping or musical-noise artefacts. Trained with a large detail-preservation loss so breaths, plosives and vocal texture stay intact.
Variants
Two variants ship. Their Core ML artifacts share the exact planar
spec / feat_erb / feat_spec → spec_enhanced I/O contract and use a fixed
batch of four independent two-second chunks. The ONNX artifacts retain their
original DFN3 layout.
clear-studio
The default. Quiet, studio-like character; silences sit close to true zero.
Best for solo podcasts, tutorials, voiceover, video demos, screen recordings, and anything that wants a clean broadcast feel.
| File | Purpose | Size |
|---|---|---|
clear-studio.mlmodelc |
ANE-optimized Core ML (fp16 compute + 6-bit weight palette, iOS 16 target) | 9.0MB |
clear-studio.mlmodelc.zip |
Same compiled model, zipped | 8.6MB |
clear-studio.onnx |
Android / cross-platform ONNX (fp16 weights, fp32 I/O) | 24MB |
clear-studio.pt |
PyTorch checkpoint, for research and re-export | 46MB |
clear-natural
Preserves room tone, breath, and lip texture.
For treated podcast studios, intentional voiceover, interviews where the room is part of the take, and remote guest recordings where absolute silence would sound wrong.
| File | Purpose | Size |
|---|---|---|
clear-natural.mlmodelc |
ANE-optimized Core ML (fp16 compute + 6-bit weight palette, iOS 16 target) | 9.0MB |
clear-natural.mlmodelc.zip |
Same compiled model, zipped | 8.6MB |
clear-natural.onnx |
Android / cross-platform ONNX | 24MB |
clear-natural.pt |
PyTorch checkpoint | 46MB |
Performance
The Core ML variants are optimized for the Apple Neural Engine.
MLComputePlan confirms that all 492 model operations run on ANE.
clear-studio, whole SDK pipeline on a 60-second clip, best of three:
| Device | Realtime factor |
|---|---|
| iPhone 16 Pro | 302x |
| MacBook Pro (M5) | 345x |
On iPhone 16 Pro, first-ever model loading takes approximately 3.4 seconds while Core ML compiles the ANE program. Cached launches load in approximately 62ms; applications should warm the model in the background.
Try it
| Platforms | iOS, macOS, tvOS, visionOS, Android, Linux, Windows, Browser, Node |
| Weights | v0.3.0 |
Install
Swift (requirements)
.package(url: "https://github.com/Desert-Ant-Labs/desert-ant-core.git", from: "3.1.0")
Then add the Clear product to your target.
Kotlin (requirements)
implementation("ai.desertant:clear:3.1.0")
JavaScript (requirements)
npm i @desert-ant-labs/clear @litertjs/core # browser
npm i @desert-ant-labs/clear # Node, prebuilt native core
What it's good for
- Meeting recorders. Zoom, Teams, Meet, Detail exports, single or multi-speaker.
- Bluetooth microphones. AirPods, Sony, headset mics.
- Mobile devices. iPhone and Android built-in microphone recordings, voice notes, field recordings.
- Laptop built-in microphones. MacBook and PC built-in mics.
- Untreated rooms. Bedrooms, hotel rooms, kitchens, coffee shops.
Whenever the pitch is messy recording in, clean audio out.
What it is not
- Not a general-purpose audio denoiser. Speech is the target; music, effects, and non-vocal signals get pulled down as noise.
- Not a source separator. Overlapping speakers stay overlapping.
- Not a voice changer, cloner, or transcription model.
Keywords
speech enhancement · noise suppression · dereverberation · speech denoising · reduce noise · clean up audio · normalize volume · turn a recording into studio sound · messy recording in clean audio out · podcast audio · voice cleanup · meeting recorder cleanup · bluetooth microphone cleanup · mobile device audio · built-in microphone · on-device audio · edge ML · Core ML · ONNX · iOS speech enhancement · Android speech enhancement · real-time speech enhancement · DFN3 · DeepFilterNet · studio sound · podcast sound · TCN · distilled model · Apple Neural Engine · ANE
License
Desert Ant Labs Source-Available License. Free for most apps, and a commercial license is required at scale. Full terms are at the link. Licensing: licensing@desertant.com.
Citation
@software{clear_2026,
title = {Clear: On-device speech enhancement: denoise, dereverb, and loudness-normalize},
author = {Desert Ant Labs},
year = {2026},
url = {https://huggingface.co/desert-ant-labs/clear},
}
© 2026 Desert Ant Labs · https://desertant.com
- Downloads last month
- 350