Guaraní and Jopara conversational speech: an anonymous source specification

What this source actually describes

The source describes adult, unscripted speech and a commercial licensing route. Those statements are useful qualification leads, but they do not establish a verified transcript corpus, a speaker count or the exact language mixture. In particular, a language name and a code-switching label should be grounded in the proposed recordings and the annotation policy used for each utterance.

This public source brief uses an anonymous reference so the conversation can start without publishing a seller identity, buyer identity, private contact or signed agreement. The reference identifies this sourcing discussion, not a certification, a reserved quantity or a completed transaction.

What is verified, reported or still open

No TTS, voice-cloning, speaker-identification or unrestricted redistribution permission is inferred from a general commercial-license description. Acoustic format, overlap, noise, recording channels and transcript accuracy remain to be inspected.

A source-reported fact is labeled as such above. A documented total, an inspected media property and a completed delivery are different kinds of evidence. Fiund confirms the current source, selection and permitted use for the actual project rather than treating a catalog description as a transferable license.

Inspect a representative pilot before scaling

Select a bounded consent-reviewed recording sample before choosing a large volume target. Agree what counts as Guaraní, Jopara or an unresolved label for the actual task. Preserve original reference text alongside any normalized scoring version, and make disagreements reviewable.

• Verify language labels at the required recording or utterance level with qualified reviewers.
• Document the transcript convention, orthography choices and treatment of mixed-language speech.
• Review adult participation, collection notices, consent scope and the authority to license the recordings.
• Separate speakers and sessions appropriately when constructing evaluation groups.

Define the task and accepted deliverable

The relevant project family is speech recognition, conversational understanding and explicitly scoped code-switching evaluation. The scope should name the unit a buyer will evaluate, the source material available to the model and the reference evidence needed to judge an output.

A source-specific audio/transcript pilot, language-label policy and speaker/session grouping specification if the evidence supports that scope.

A proposal should distinguish source access, selection, preparation, labeling and acceptance work. There is no published blanket price for this source. Counting files, hours, projects, transcripts or capture-system cameras as though they were the same accepted unit would conceal the actual work.

Request the source through an anonymous reference

When contacting Fiund, use reference SP-001 and describe the intended model task, required source characteristics, target quantity and evaluation conditions. Start with project requirements rather than confidential data or participant information.

Fiund can review which source evidence can be shared for that discussion and which permissions or agreements must be completed before access. This page does not name a client, imply an endorsement, guarantee the source remains available or grant a license to the underlying material.

Questions buyers ask