Browse categories
Explore
Fiverr Pro
English
$
USD
I will record pashto or dari speech dataset for ai model training
Mansour Sadat
About this gig
Are you looking for clean, native Pashto or Persian/Dari voice recordings for your AI, speech recognition, or NLP projects?
You're in the right place!
I will record high-quality utterances in Pashto or Dari with native accuracy, perfect for training AI models, speech-to-text (STT), and ASR systems.
I will provide you with:
- Noise-free WAV/MP3 audio
- Native Pashto & Persian/Dari accents
- Transcription + labeling + metadata (CSV/Excel)
- Multiple speakers available (on request)
Whether you need a small starter dataset or a large-scale speech corpus, I can deliver fast, reliable, and professionally proof-listened recordings.
Let's build your AI dataset with clarity, accuracy, and trust!
Get to know Mansour Sadat
Mansour Sadat
Innovative Frontend Web Developer and Fluent Trilingual Translator
- FromAfghanistan
- Member sinceJul 2024
- Avg. response time1 hour
Languages
English, Pashto, Persian
I'm Sayed Mansour Sadat, a Front-End Developer, AI Data Specialist, and Trilingual Language Professional (English, Dari/Farsi, and Pashto) with more than three years of experience.
I specialize in AI evaluation, language-related tasks, data annotation, and multilingual workflows while also building clean, responsive websites. I focus on accuracy, clear communication, fast delivery, and delivering reliable results for every project.
My Portfolio
FAQ
What exactly do you provide in the recordings?
I provide Pashto or Persian/Dari utterances in clean, noise-free audio (WAV/MP3). Depending on your package, I also include transcription, labeling, and metadata in CSV/Excel format.
What is metadata?
Metadata is structured information about each audio file (e.g., filename, utterance text, speaker ID, duration). This makes your dataset easy to organize and use for AI/ML projects.
What is transcription?
Transcription is the written text version of the audio recordings, useful for training speech recognition models.
What is labeling?
Labeling means tagging the dataset (e.g., by speaker, gender, utterance type, or category) so AI models can recognize patterns more effectively.
Can I request multiple speakers?
Yes! By default, I record with one native speaker, but you can order the “Additional Speaker” gig extra for more voices.
What if I need a larger dataset (thousands of utterances)?
Please send me a custom order — I can scale up and create a tailored dataset for your project by participation of numerous native speakers.
In what formats do you deliver?
Audio is delivered in WAV or MP3, while text/transcription/metadata is delivered in TXT, CSV, or Excel, based on your preference.
