Multilingual Speech Collection & Annotation
Budget / Salary₹12,500–37,500
TypeFreelance project
LocationRemote
Posted1 hour ago
I’m expanding Project 510 and need fresh, conversational speech recordings that I can feed directly into a new voice-assistant model. Your task is two-fold: capture clean audio and attach accurate speaker metadata. The recordings can be spontaneous dialogue, read prompts, or any natural speech, but every file must be tagged with the speaker’s language, plus accent or regional style and age group. The language label is the single most critical field, so consistency there is non-negotiable.
You may recruit contributors, use an existing panel, or supply licensed corpora you already own—as long as usage rights transfer to me. Technical specs are straightforward: 44.1 kHz, 16-bit WAV (or higher), minimal background noise, no clipping. I’ll provide a simple template for logging the metadata and naming files.
Deliverables (per batch)
• High-quality WAV files
• Completed metadata sheet with language, accent/region, age group for each speaker
Acceptance criteria
• Audio meets format and clarity specs
• 99 % accuracy on language tags
• Metadata sheet fully populated and cross-checked against files
Tell me which languages you can cover, how many hours of speech you can supply each week, and your QC process. Ready when you are—let’s build a great dataset together.
You may recruit contributors, use an existing panel, or supply licensed corpora you already own—as long as usage rights transfer to me. Technical specs are straightforward: 44.1 kHz, 16-bit WAV (or higher), minimal background noise, no clipping. I’ll provide a simple template for logging the metadata and naming files.
Deliverables (per batch)
• High-quality WAV files
• Completed metadata sheet with language, accent/region, age group for each speaker
Acceptance criteria
• Audio meets format and clarity specs
• 99 % accuracy on language tags
• Metadata sheet fully populated and cross-checked against files
Tell me which languages you can cover, how many hours of speech you can supply each week, and your QC process. Ready when you are—let’s build a great dataset together.
Apply on Freelancer →
Project sourced from Freelancer.com. Applications happen directly on the original platform — we never collect your data.