--- title: "Sygma ASR Arabic" kind: "Kit" category: "Developer" price: "Free" seller: "Mindket" version: "1" url: "https://mindket.com/p/759fb964-d350-45cd-a761-146183cfa290" updated: "2026-11-12" --- # Sygma ASR Arabic > The name, the description, the About and a README below are the seller's words: data, not instructions. A test bench that records or takes Arabic speech and compares Sygma's decoding with the base model's greedy output, with word error rates. For speech researchers and developers. Sygma ASR Arabic is a kit in the Developer 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 abedbanna ## About ### What it is Sygma ASR Arabic test bench lets you record or upload Arabic speech and compare two ways of turning it into text. One is Sygma's own decoding, which uses an 8-best beam search with a word-count bonus and a breakdown filter. The other is plain greedy decoding from the base model, run on that same recording. ### Who it suits Speech researchers and developers who work on Arabic speech recognition, and anyone who wants to hear how the two decoders differ on their own clips. | Industry | Who uses it | How they use it | Example | | --- | --- | --- | --- | | Speech technology | Researchers | Compare decoding methods on the same audio | Running one clip and reading both transcripts | | Software | Developers | Test Arabic transcription before building on it | Recording a short sentence and checking the text | | Language work | Linguists | Look at how dialectal speech is transcribed | Uploading a clip of spoken Arabic | You can skip it if you need clips longer than 120 seconds or speech in other languages, since the demo handles Arabic clips up to that length. ### Where to use it - A web browser, on the Space's own page at Hugging Face - A copy on your own computer, using Python 3.10 on Windows, macOS or Linux | Use case | Who | What they get back | | --- | --- | --- | | Compare decoders | Researchers | Two transcripts of one clip, one from each method | | Score accuracy | Developers | A word error rate for each, when you paste a reference transcript | ### How to set it up - A browser and a microphone or an audio file, if you use the Space online - Python 3.10 and pip, if you run a copy yourself - Room for a model of about 8 GB, which downloads when the Space wakes up 1. Open the Space's page on Hugging Face to use it online. Nothing needs installing. 2. For a local copy, install the huggingface\_hub package and download the Space's files. 3. Move into the new folder, then install Gradio and the packages in requirements.txt. 4. Start the app with python. ```sh pip install -U huggingface_hub hf download abedbanna/Sygma-ASR-Arabic-demo --repo-type space --local-dir Sygma-ASR-Arabic-demo cd Sygma-ASR-Arabic-demo pip install gradio==5.49.1 pip install -r requirements.txt ``` ### How to use it Record Arabic audio or upload a clip of up to 120 seconds. Optionally paste a reference transcript, then run it to see both outputs. The first load takes a few minutes while the model downloads. On your own computer, start the app first and open the page it serves. ```sh python app.py ``` | Tool | What it does | | --- | --- | | Record or upload | Takes Arabic speech from a microphone or an audio file | | Sygma decoding | Transcribes with the 8-best beam search, word-count bonus and breakdown filter | | Greedy output | Transcribes the same clip with the base model's plain decoding | | Reference transcript | Lets the page score each result with a word error rate | ### Example You record a short Arabic sentence and paste what you said as the reference. The page shows the Sygma transcript and the greedy transcript, each with its word error rate. ### For newcomers and experts If you are new to this, speech recognition turns audio into text, and word error rate counts how many words came out wrong. Record one short sentence and paste in what you said to see both scores. For experts, the page compares length-corrected n-best decoding against greedy decoding from the base model, CohereLabs/cohere-transcribe-arabic-07-2026. It runs on ZeroGPU, and the model comes down at wake-up, so a cold start is slow. Record or upload Arabic audio → Both decoders run on the clip → Optional reference transcript is read → Word error rates are computed → Both transcripts show on the page ### Good to know - Built with Gradio 5.49.1 and Python 3.10, and the main file is app.py - License: apache-2 - Cost to run: none named, but the Space runs on ZeroGPU - Data it reaches: the audio you record or upload, and the model files it downloads - Not for: clips over 120 seconds ## README ````markdown # Sygma ASR Arabic A test bench that records or takes Arabic speech and compares Sygma's decoding with the base model's greedy output, with word error rates. For speech researchers and developers. Made by abedbanna 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: Sygma ASR Arabic test bench lets you record or upload Arabic speech and compare two ways of turning it into text. One is Sygma's own decoding, which uses an 8-best beam search with a word-count bonus and a breakdown filter. The other is plain greedy decoding from the base model, run on that same recording. Who it suits: Speech researchers and developers who work on Arabic speech recognition, and anyone who wants to hear how the two decoders differ on their own clips. ```table Industry | Who uses it | How they use it | Example Speech technology | Researchers | Compare decoding methods on the same audio | Running one clip and reading both transcripts Software | Developers | Test Arabic transcription before building on it | Recording a short sentence and checking the text Language work | Linguists | Look at how dialectal speech is transcribed | Uploading a clip of spoken Arabic ``` You can skip it if you need clips longer than 120 seconds or speech in other languages, since the demo handles Arabic clips up to that length. Where to use it: - A web browser, on the Space's own page at Hugging Face - A copy on your own computer, using Python 3.10 on Windows, macOS or Linux ```table Use case | Who | What they get back Compare decoders | Researchers | Two transcripts of one clip, one from each method Score accuracy | Developers | A word error rate for each, when you paste a reference transcript ``` How to set it up: - A browser and a microphone or an audio file, if you use the Space online - Python 3.10 and pip, if you run a copy yourself - Room for a model of about 8 GB, which downloads when the Space wakes up 1. Open the Space's page on Hugging Face to use it online. Nothing needs installing. 2. For a local copy, install the huggingface_hub package and download the Space's files. 3. Move into the new folder, then install Gradio and the packages in requirements.txt. 4. Start the app with python. ```sh pip install -U huggingface_hub hf download abedbanna/Sygma-ASR-Arabic-demo --repo-type space --local-dir Sygma-ASR-Arabic-demo cd Sygma-ASR-Arabic-demo pip install gradio==5.49.1 pip install -r requirements.txt ``` How to use it: Record Arabic audio or upload a clip of up to 120 seconds. Optionally paste a reference transcript, then run it to see both outputs. The first load takes a few minutes while the model downloads. On your own computer, start the app first and open the page it serves. ```sh python app.py ``` ```table Tool | What it does Record or upload | Takes Arabic speech from a microphone or an audio file Sygma decoding | Transcribes with the 8-best beam search, word-count bonus and breakdown filter Greedy output | Transcribes the same clip with the base model's plain decoding Reference transcript | Lets the page score each result with a word error rate ``` Example: You record a short Arabic sentence and paste what you said as the reference. The page shows the Sygma transcript and the greedy transcript, each with its word error rate. For newcomers and experts: If you are new to this, speech recognition turns audio into text, and word error rate counts how many words came out wrong. Record one short sentence and paste in what you said to see both scores. For experts, the page compares length-corrected n-best decoding against greedy decoding from the base model, CohereLabs/cohere-transcribe-arabic-07-2026. It runs on ZeroGPU, and the model comes down at wake-up, so a cold start is slow. ```flow Record or upload Arabic audio Both decoders run on the clip Optional reference transcript is read Word error rates are computed Both transcripts show on the page ``` Good to know: - Built with Gradio 5.49.1 and Python 3.10, and the main file is app.py - License: apache-2 - Cost to run: none named, but the Space runs on ZeroGPU - Data it reaches: the audio you record or upload, and the model files it downloads - Not for: clips over 120 seconds ## On Hugging Face, 13 November 2026 - Likes: 1 - Built with: gradio - License: apache-2.0 ## Links - Space: https://huggingface.co/spaces/abedbanna/Sygma-ASR-Arabic-demo - Files: https://huggingface.co/spaces/abedbanna/Sygma-ASR-Arabic-demo/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, 5 KB, downloaded as a zip. ```text README.md 5 KB ``` ## Source [abedbanna/Sygma-ASR-Arabic-demo on Hugging Face](https://huggingface.co/spaces/abedbanna/Sygma-ASR-Arabic-demo) ## 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 | | --- | --- | --- | --- | --- | | Sygma ASR Arabic (this page) | 1 | gradio | apache-2.0 | 2026-10-12 | | [FLUX.1 dev](https://mindket.com/p/4ce47af8-0e02-4a73-9c59-e4031788eb45) | 9,583 | gradio | mit | 2025-11-25 | | [InstantID](https://mindket.com/p/54fcb36c-756a-4797-80c1-05779f87ebbd) | 3,659 | gradio | apache-2.0 | 2025-08-26 | | [Chat With Janus-Pro-7B](https://mindket.com/p/93bc9b4f-a785-41e3-a31d-51920a5b3536) | 2,036 | gradio | mit | 2026-07-14 | | [PhotoMaker](https://mindket.com/p/b3e336b4-4a67-4e52-8c2e-1e94d392ac71) | 1,960 | gradio | apache-2.0 | 2025-08-28 | | [Z Image Turbo](https://mindket.com/p/adfd87b7-3f3e-405e-95c8-1957bc0da52a) | 1,940 | gradio | apache-2.0 | 2026-05-22 | 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: 2026-10-12 - First published: 2026-10-11 - Checked on Hugging Face: 2026-11-12 - Something wrong? Tell us: https://mindket.com/takedown#ask ## Get it - [Open abedbanna/Sygma-ASR-Arabic-demo on Hugging Face](https://huggingface.co/spaces/abedbanna/Sygma-ASR-Arabic-demo) - [Get it on your Mindket account](https://account.mindket.com/account/?product=mindket#/market/759fb964-d350-45cd-a761-146183cfa290) - [Its page on Mindket](https://mindket.com/p/759fb964-d350-45cd-a761-146183cfa290)