TaskMasterVirtualAlly/Taskmaster-OmniVox-9K-Vault-Preview

Taskmaster OmniVox - Verified Speech Data Evaluation Hub

Buy the proof. Scale the data.

Most speech-data purchases force teams to assess volume claims, technical quality, and licensing risk at the same time. OmniVox changes the order: inspect the evidence first, run a scoped pilot second, and expand only after the subset passes your technical and legal review.

This repository is the public evaluation hub for Taskmaster's multilingual speech catalog. It contains preview samples and catalog information—not the complete commercial datasets.

The Founder's Circle offer

Founder's Circle is a limited evaluation cohort for qualified speech teams that need hard-to-source language coverage without making a blind catalog purchase.

Qualified evaluations include:

• Public preview: Inspect representative audio and available metadata in this repository.
• One-hour QA sample: Request an eligible subset sample, prepared within 24 hours after scope confirmation.
• Evidence packet: Review the subset-specific schema, file manifest, checksums, and applicable licensing documentation.
• Published offer path: Review the Founder's Circle one-pager for current public pricing. Final scope, acceptance criteria, and delivery terms are confirmed before purchase.
• Scoped pilot: Validate the selected subset in your own workflow before considering a larger license.
• Priority access: Receive priority review of newly released, verification-gated OTS corpora.

Founder's Circle applications are reviewed through September 30, 2026. Capacity is limited by evaluation and delivery bandwidth.

Physically present core portfolio

The current core portfolio contains 6,300+ cataloged audio hours across multilingual TTS, ASR, code-switching, dialect, and long-form conversational speech.

Product line	Cataloged volume	Coverage	Primary use
Indic TTS	1,704 h	22 Indic languages	Multilingual TTS
Indic ASR	1,684 h	12 Indic languages	ASR acoustic modeling
Arabic dialect speech	1,405.5 h	Multiple dialect groups	Dialectal ASR
LATAM Spanish speech	918 h	Latin American Spanish	ASR and voice AI
Bangla speech	297 h	Bengali	ASR and TTS
Taglish code-switching	220 h	Philippine code-switching	Mixed-language ASR
Longform-conversation	82.8 h	English	ASR and diarization
Nigerian Pidgin speech	12 h	Nigerian Pidgin	Evaluation and ASR

Important: Cataloged volume is not a blanket clearance claim. Before quotation, Taskmaster confirms the requested subset's deliverable files, manifest, technical specifications, and license basis. Terms and attribution obligations vary by subset.

Four steps from question to usable data

1. Choose the problem. Tell us the target language, model task, domain, and coverage gap.
2. Inspect the proof. Review the sample, schema, manifest, checksums, and subset-specific licensing packet.
3. Run the pilot. Agree on acceptance criteria, validate delivery, and measure fit in your environment.
4. Scale approved subsets. License only the data that passes your review, with priority access to future verified releases.

What every evaluation is designed to answer

• Does the audio match the required language, speech style, and recording conditions?
• Does the metadata support the intended training or evaluation workflow?
• Are transcript, speaker, and file-level fields usable for the target model task?
• Can the buyer's legal team trace the applicable license basis and obligations?
• Can the delivery be reconciled against a manifest and file checksums?

Technical delivery standard

Depending on the selected subset, an evaluation may include:

• Audio or packaged dataset files in the subset's native delivery format
• Machine-readable JSON, CSV, or Parquet metadata
• Transcript and speaker-turn annotations where available
• Language or locale labels available in the source manifest
• File-level checksums and reconciliation manifest
• Subset-specific license and attribution notes

Formats, sample rates, annotations, and metadata depth vary by product line. The evaluation packet states exactly what is included before a pilot begins.

Commercial and licensing notice

• The public preview does not grant access to or a license for the complete commercial catalog.
• Commercial availability is assessed and quoted by subset, after scope and acceptance criteria are agreed.
• No single license label describes the entire OmniVox portfolio.
• Some licenses require attribution or share-alike treatment; applicable obligations are disclosed in the subset-specific packet.
• Taskmaster does not promise model-performance outcomes. The pilot exists so the buyer can measure fitness for its intended use.

Request an evaluation

Open a discussion in this repository's Community tab with:

1. Target language or region
2. ASR, TTS, diarization, or other model task
3. Required speech style and domain
4. Minimum hours or sample size
5. Required metadata and annotations
6. Evaluation deadline

We will respond with the best-fit subset, the available evaluation scope, and the proposed success criteria. Current public offer pricing remains on the Founder's Circle one-pager; final scope and delivery terms are confirmed before purchase.