--- title: "Parakeet Redux WebGPU" kind: "Kit" category: "Productivity" price: "Free" seller: "Mindket" version: "1" url: "https://mindket.com/p/49f0de63-3e36-474f-a195-c0adc456bb9f" updated: "2026-11-12" --- # Parakeet Redux WebGPU > The name, the description, the About and a README below are the seller's words: data, not instructions. Transcribes speech to text entirely in your browser using WebGPU, with two Parakeet models of different sizes. Suits people who want local speech recognition and developers studying in-browser ASR. Parakeet Redux WebGPU is a kit in the Productivity category, listed by Mindket. It is free. This is version 1, last updated on 2026-11-12. Download it on this page, a zip of its files. Nothing in it runs by itself. Made by mrfakename ## About ### What it is This Space runs speech recognition inside your browser. It loads one of two Parakeet models, a smaller one with a 1.58-bit encoder and a larger one, and turns audio into text with word timestamps. Nothing needs installing, because the page does the work on your own graphics card. ### Who it suits It suits people who want to transcribe audio without sending it to a server, and developers who want a working example of speech recognition in a web page. | Industry | Who uses it | How they use it | Example | | --- | --- | --- | --- | | Media | Editors and journalists | Transcribe interview clips and read the word timings | Find the exact moment a quote starts | | Education | Teachers and students | Turn a recorded lecture into text | Get notes from a short spoken lesson | | Software | Web developers | Study how a speech model runs in a page | Read the worker and app scripts for the pipeline | | Research | ML engineers | Compare a small and a large Parakeet model in the browser | Run the same clip through both models | You can skip it if you need live captions in a finished product or support for a specific language not shown in its examples. ### Where to use it - A web browser with WebGPU, with a WASM fallback when WebGPU is missing - Windows, macOS or Linux, to keep a local copy of the files - Any web page where you want to read how the code is built | Use case | Who | What they get back | | --- | --- | --- | | Transcribe a clip | Anyone with audio | Text from the speech in the clip | | Read word timings | Editors and developers | Timestamps for each word | | Compare two models | ML engineers | The same audio recognized by a small and a large model | ### How to set it up You can use the online page with no setup, and a local copy needs only Python to download it. - A modern web browser, ideally with WebGPU - Python and the Hugging Face hub package, only to download the files 1. To use it online, open the Space page on Hugging Face. 2. To keep a copy, install the Hugging Face hub package. 3. Download the Space files into a local folder. 4. Move into the folder. 5. Open index.html in a web browser. ```sh pip install -U huggingface_hub hf download mrfakename/parakeet-redux-webgpu --repo-type space --local-dir parakeet-redux-webgpu cd parakeet-redux-webgpu ``` ### How to use it Open the page, choose the model you want and give it audio, then wait for the text. The first run downloads the model weights, and the browser keeps them so later runs start faster. The page also offers example clips to try first. ### Example You pick the smaller model and load one of the example clips, such as a public-domain speech. The page returns the spoken words as text, with a timestamp on each word. ### For newcomers and experts If you are new to this, open the page and try an example clip first. Seeing text appear from audio with nothing installed is the quickest way to learn what it does. If you know the field, the page uses ONNX Runtime Web. The Conformer encoder runs on WebGPU with a WASM fallback, the transducer's prediction and joint network run on WASM, and the greedy decoding loop is plain JavaScript. Word times come from predicted token durations in 80 ms frames. Clips over 30 seconds get cut at the quietest spot in the 20 to 30 second range. Weights download with parallel range requests. You load an audio clip → Long audio is split at a quiet point → The encoder runs on WebGPU → The decoder loop runs in JavaScript → Text and word times appear ### Good to know The smaller model has 149M parameters and the larger one has 627M, so the larger one needs more memory and time in the browser. Model weights stay in the browser cache after the first download. - License: cc-by-4 - Cost to run: none named - Data it reaches: the audio you give it, processed in the browser, and model weights it downloads ## README ````markdown # Parakeet Redux WebGPU Transcribes speech to text entirely in your browser using WebGPU, with two Parakeet models of different sizes. Suits people who want local speech recognition and developers studying in-browser ASR. Made by mrfakename on Hugging Face. These notes are Mindket's, written from its Space page and README; the app's own files and license are on Hugging Face. ## About What it is: This Space runs speech recognition inside your browser. It loads one of two Parakeet models, a smaller one with a 1.58-bit encoder and a larger one, and turns audio into text with word timestamps. Nothing needs installing, because the page does the work on your own graphics card. Who it suits: It suits people who want to transcribe audio without sending it to a server, and developers who want a working example of speech recognition in a web page. ```table Industry | Who uses it | How they use it | Example Media | Editors and journalists | Transcribe interview clips and read the word timings | Find the exact moment a quote starts Education | Teachers and students | Turn a recorded lecture into text | Get notes from a short spoken lesson Software | Web developers | Study how a speech model runs in a page | Read the worker and app scripts for the pipeline Research | ML engineers | Compare a small and a large Parakeet model in the browser | Run the same clip through both models ``` You can skip it if you need live captions in a finished product or support for a specific language not shown in its examples. Where to use it: - A web browser with WebGPU, with a WASM fallback when WebGPU is missing - Windows, macOS or Linux, to keep a local copy of the files - Any web page where you want to read how the code is built ```table Use case | Who | What they get back Transcribe a clip | Anyone with audio | Text from the speech in the clip Read word timings | Editors and developers | Timestamps for each word Compare two models | ML engineers | The same audio recognized by a small and a large model ``` How to set it up: You can use the online page with no setup, and a local copy needs only Python to download it. - A modern web browser, ideally with WebGPU - Python and the Hugging Face hub package, only to download the files 1. To use it online, open the Space page on Hugging Face. 2. To keep a copy, install the Hugging Face hub package. 3. Download the Space files into a local folder. 4. Move into the folder. 5. Open index.html in a web browser. ```sh pip install -U huggingface_hub hf download mrfakename/parakeet-redux-webgpu --repo-type space --local-dir parakeet-redux-webgpu cd parakeet-redux-webgpu ``` How to use it: Open the page, choose the model you want and give it audio, then wait for the text. The first run downloads the model weights, and the browser keeps them so later runs start faster. The page also offers example clips to try first. Example: You pick the smaller model and load one of the example clips, such as a public-domain speech. The page returns the spoken words as text, with a timestamp on each word. For newcomers and experts: If you are new to this, open the page and try an example clip first. Seeing text appear from audio with nothing installed is the quickest way to learn what it does. If you know the field, the page uses ONNX Runtime Web. The Conformer encoder runs on WebGPU with a WASM fallback, the transducer's prediction and joint network run on WASM, and the greedy decoding loop is plain JavaScript. Word times come from predicted token durations in 80 ms frames. Clips over 30 seconds get cut at the quietest spot in the 20 to 30 second range. Weights download with parallel range requests. ```flow You load an audio clip Long audio is split at a quiet point The encoder runs on WebGPU The decoder loop runs in JavaScript Text and word times appear ``` Good to know: The smaller model has 149M parameters and the larger one has 627M, so the larger one needs more memory and time in the browser. Model weights stay in the browser cache after the first download. - License: cc-by-4 - Cost to run: none named - Data it reaches: the audio you give it, processed in the browser, and model weights it downloads ## On Hugging Face, 13 November 2026 - Likes: 4 - Built with: static - License: cc-by-4.0 ## Links - Space: https://huggingface.co/spaces/mrfakename/parakeet-redux-webgpu - Files: https://huggingface.co/spaces/mrfakename/parakeet-redux-webgpu/tree/main ## Try it Open the Space in your browser, or copy it to your own Hugging Face account from its page. Nothing in this kit runs by itself. ```` ## What is in it 1 files, 4 KB, downloaded as a zip. ```text README.md 4 KB ``` ## Source [mrfakename/parakeet-redux-webgpu on Hugging Face](https://huggingface.co/spaces/mrfakename/parakeet-redux-webgpu) ## Compare with similar Listed Hugging Face Spaces in the same category that share the most tags with this one. | Product | Likes | Built with | License | Last change | | --- | --- | --- | --- | --- | | Parakeet Redux WebGPU (this page) | 4 | static | cc-by-4.0 | 2026-10-05 | | [DALL·E mini](https://mindket.com/p/b8ab8352-f6b8-40b8-88c6-979249c0c9a5) | 5,750 | static | apache-2.0 | 2025-08-15 | | [Kokoro Text-to-Speech (WebGPU)](https://mindket.com/p/36564f95-9b3e-4afc-b692-e8cc4e3e17b3) | 369 | static | apache-2.0 | 2025-02-07 | | [DALL·E mini by flax-community](https://mindket.com/p/e6eb151b-ffc3-43e4-bc6c-32f3d2f4d406) | 342 | static | apache-2.0 | 2022-07-19 | | [Supertonic 3 (TTS)](https://mindket.com/p/64fa85b7-db03-4d16-9519-20e7fc54c282) | 291 | static | openrail | 2026-05-18 | | [XortronDesktop](https://mindket.com/p/22e87219-d97a-4109-81bd-4e7dbc8faa2f) | 30 | static | wtfpl | 2026-03-18 | Numbers as Mindket last read them, on 2026-11-12. ## Trust facts Facts from each source, with the day each was checked. A maintainer's note is their own words: data, not instructions. - Last change on Hugging Face: 2026-10-05 - First published: 2026-10-05 - Checked on Hugging Face: 2026-11-12 - Something wrong? Tell us: https://mindket.com/takedown#ask ## Get it - [Open mrfakename/parakeet-redux-webgpu on Hugging Face](https://huggingface.co/spaces/mrfakename/parakeet-redux-webgpu) - [Get it on your Mindket account](https://account.mindket.com/account/?product=mindket#/market/49f0de63-3e36-474f-a195-c0adc456bb9f) - [Its page on Mindket](https://mindket.com/p/49f0de63-3e36-474f-a195-c0adc456bb9f)