--- title: "whisper-persian-v4" kind: "Kit" category: "Developer" price: "Free" seller: "Mindket" version: "1" url: "https://mindket.com/p/e05b0e6b-d638-4f8e-a01f-88dd76669fc5" updated: "2026-11-13" --- # whisper-persian-v4 > The name, the description, the About and a README below are the seller's words: data, not instructions. Turns Persian speech into text. It is a fine-tune of Whisper large v3 on a Persian dataset, for developers and teams that transcribe Persian audio. whisper-persian-v4 is a kit in the Developer category, listed by Mindket. It is free. This is version 1, last updated on 2026-11-13. Download it on this page, a zip of its files. Nothing in it runs by itself. Made by nezamisafa ## About ### What it is whisper-persian-v4 is a speech recognition model that writes Persian audio out as text. It is a fine-tuned version of OpenAI's Whisper large v3, trained on a Persian dataset named ASR\_fa\_v1. The card lists a word error rate of 8.73 on its evaluation set. ### Who it suits It suits developers, media teams and researchers who need Persian speech turned into text. The card is short on detail, so expect to test it on your own recordings. | Industry | Who uses it | How they use it | Example | | --- | --- | --- | --- | | Media | Editors and podcasters | Transcribe Persian recordings for captions or notes | A producer turns an interview recording into editable text | | Software | App developers | Add Persian speech to text inside an app | A developer adds voice input for Persian speakers | | Research | Speech researchers | Compare its error rate with other Persian models | A researcher runs it on a test set and records the errors | If your audio is in another language, you can skip it, since the record lists Persian only. ### Where to use it - Python with the transformers library - Windows, macOS or Linux, after downloading the files with Python | Use case | Who | What they get back | | --- | --- | --- | | Transcribe a recording | Editor or developer | Persian text for the spoken words | | Voice input in an app | Developer | Text from a user's speech | | Benchmark a Persian model | Researcher | Transcripts to score against reference text | ### How to set it up - Python on your computer - The transformers library - Disk space for about 6.2 GB in F32, or about 772 MB at 4-bit by Hugging Face's estimate 1. Install the Hugging Face hub package. 2. Download the model files. 3. Load them with transformers. The card shows no code sample, so follow the transformers documentation for speech recognition models. ```sh pip install -U huggingface_hub hf download nezamisafa/whisper-persian-v4 ``` ### How to use it You pass the model a Persian audio recording and it returns the words as text. The only commands the source shows are the download steps. ```sh hf download nezamisafa/whisper-persian-v4 ``` | Tool | What it does | | --- | --- | | hf download | Fetches the model files to your computer | | pip install -U huggingface\_hub | Installs the package that provides the hf command | ### Example You give it a short voice memo spoken in Persian. The model returns the spoken words as Persian text, which you can then proofread and use as captions or notes. ### For newcomers and experts A speech recognition model listens to audio and writes down what it hears. This one was taught on Persian. A first small thing to try is a short, clear recording of one speaker, to see how close the text is to what was said. For experts, it fine-tunes Whisper large v3 with 1.5B parameters. The card reports training for 5000 steps at a learning rate of 1e-05, with a batch size of 16 and mixed precision. Word error rate fell from 17.1059 at step 1000 to 8.7300 at step 5000. The card gives no description of the training data beyond its name. OpenAI Whisper large v3 is the base → It is fine-tuned on the ASR\_fa\_v1 dataset → You download the files with hf → transformers loads the model → Persian audio goes in and text comes out ### Good to know - Gaps: the intended uses, limits and training data sections of the card say more information is needed - Language: Persian - License: Apache 2.0 - Cost to run: none named - Data it reaches: downloading uses the web and writes files to your disk, and the audio you feed it is the only other input - Not for: anything the card rules out, since it names no limits, so test on your own audio before relying on it ## README ````markdown # whisper-persian-v4 Turns Persian speech into text. It is a fine-tune of Whisper large v3 on a Persian dataset, for developers and teams that transcribe Persian audio. Made by nezamisafa on Hugging Face. These notes are Mindket's, written from its model page and model card; the model's own files and license are on Hugging Face. ## About What it is: whisper-persian-v4 is a speech recognition model that writes Persian audio out as text. It is a fine-tuned version of OpenAI's Whisper large v3, trained on a Persian dataset named ASR_fa_v1. The card lists a word error rate of 8.73 on its evaluation set. Who it suits: It suits developers, media teams and researchers who need Persian speech turned into text. The card is short on detail, so expect to test it on your own recordings. ```table Industry | Who uses it | How they use it | Example Media | Editors and podcasters | Transcribe Persian recordings for captions or notes | A producer turns an interview recording into editable text Software | App developers | Add Persian speech to text inside an app | A developer adds voice input for Persian speakers Research | Speech researchers | Compare its error rate with other Persian models | A researcher runs it on a test set and records the errors ``` If your audio is in another language, you can skip it, since the record lists Persian only. Where to use it: - Python with the transformers library - Windows, macOS or Linux, after downloading the files with Python ```table Use case | Who | What they get back Transcribe a recording | Editor or developer | Persian text for the spoken words Voice input in an app | Developer | Text from a user's speech Benchmark a Persian model | Researcher | Transcripts to score against reference text ``` How to set it up: - Python on your computer - The transformers library - Disk space for about 6.2 GB in F32, or about 772 MB at 4-bit by Hugging Face's estimate 1. Install the Hugging Face hub package. 2. Download the model files. 3. Load them with transformers. The card shows no code sample, so follow the transformers documentation for speech recognition models. ```sh pip install -U huggingface_hub hf download nezamisafa/whisper-persian-v4 ``` How to use it: You pass the model a Persian audio recording and it returns the words as text. The only commands the source shows are the download steps. ```sh hf download nezamisafa/whisper-persian-v4 ``` ```table Tool | What it does hf download | Fetches the model files to your computer pip install -U huggingface_hub | Installs the package that provides the hf command ``` Example: You give it a short voice memo spoken in Persian. The model returns the spoken words as Persian text, which you can then proofread and use as captions or notes. For newcomers and experts: A speech recognition model listens to audio and writes down what it hears. This one was taught on Persian. A first small thing to try is a short, clear recording of one speaker, to see how close the text is to what was said. For experts, it fine-tunes Whisper large v3 with 1.5B parameters. The card reports training for 5000 steps at a learning rate of 1e-05, with a batch size of 16 and mixed precision. Word error rate fell from 17.1059 at step 1000 to 8.7300 at step 5000. The card gives no description of the training data beyond its name. ```flow OpenAI Whisper large v3 is the base It is fine-tuned on the ASR_fa_v1 dataset You download the files with hf transformers loads the model Persian audio goes in and text comes out ``` Good to know: - Gaps: the intended uses, limits and training data sections of the card say more information is needed - Language: Persian - License: Apache 2.0 - Cost to run: none named - Data it reaches: downloading uses the web and writes files to your disk, and the audio you feed it is the only other input - Not for: anything the card rules out, since it names no limits, so test on your own audio before relying on it ## On Hugging Face, 13 November 2026 - Task: automatic speech recognition - Library: transformers - Parameters: 1.5B - Weights: about 6.2 GB in F32; about 772 MB at 4-bit (estimate) - Formats: Safetensors - Downloads in the last 30 days: 1,071 - Likes: 21 - License: apache-2.0 - Languages: fa - Based on: openai/whisper-large-v3 - Quantized versions: 1 - Fine-tunes: 1 - Tags: transformers, tensorboard, safetensors, whisper, automatic-speech-recognition, generated_from_trainer, model-index ## Links - Model: https://huggingface.co/nezamisafa/whisper-persian-v4 - Files: https://huggingface.co/nezamisafa/whisper-persian-v4/tree/main ## Get the model Download its files with Hugging Face's command line: ```sh pip install -U huggingface_hub hf download nezamisafa/whisper-persian-v4 ``` Then follow the usage in its model card. 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 [nezamisafa/whisper-persian-v4 on Hugging Face](https://huggingface.co/nezamisafa/whisper-persian-v4) ## Compare with similar Listed Hugging Face models in the same category that share the most tags with this one. | Product | Downloads, last 30 days | Likes | Parameters | License | Last change | | --- | --- | --- | --- | --- | --- | | whisper-persian-v4 (this page) | 1,071 | 21 | 1.5B | apache-2.0 | 2025-06-03 | | [whisper-base-ar-quran](https://mindket.com/p/58e6930e-47f6-4a82-ab75-28ef9645a961) | 353,756 | 195 | — | apache-2.0 | 2022-12-13 | | [Distil-Whisper distil-large-v3](https://mindket.com/p/d6a2c001-9ad9-472b-9889-ae7a3bee3b31) | 754,155 | 386 | 756M | mit | 2026-04-21 | | [whisper-large-v3-turbo](https://mindket.com/p/0a437e3c-069c-48a1-b07a-428a0416bd9c) | 4,969,987 | 3,524 | 809M | mit | 2024-10-04 | | [whisperkit-coreml](https://mindket.com/p/41dd2c44-8389-4b32-b641-2c39501a40dd) | 60,717,190 | 266 | — | mit | 2026-08-19 | | [Wav2Vec2-Base-960h](https://mindket.com/p/29706acd-2079-4db5-a54d-6436b4514cb3) | 1,794,378 | 418 | 94.4M | apache-2.0 | 2022-11-14 | Numbers as Mindket last read them, from 2026-11-12 to 2026-11-13. ## 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: 2025-06-03 - First published: 2025-06-02 - Checked on Hugging Face: 2026-11-13 - Something wrong? Tell us: https://mindket.com/takedown#ask ## Get it - [Open nezamisafa/whisper-persian-v4 on Hugging Face](https://huggingface.co/nezamisafa/whisper-persian-v4) - [Get it on your Mindket account](https://account.mindket.com/account/?product=mindket#/market/e05b0e6b-d638-4f8e-a01f-88dd76669fc5) - [Its page on Mindket](https://mindket.com/p/e05b0e6b-d638-4f8e-a01f-88dd76669fc5)