owensong/Inflect-v3-Preview-Media

Inflect v3 Preview Media

Public listening examples and presentation assets for the Inflect Mini, Micro, Nano and Atom v3 previews. Preview speech is not a final release or a license grant.

gallery-v2/ contains six raw 24 kHz mono clips for each preview, in this order: Cassidy, Mark, Hope, James, Sophia, Jasper. Each voice reads a distinct short text; the same six texts are used across all four models. Jasper reads a lexical stress sample. The texts were fixed before synthesis. Fifteen unchanged or same-voice matched-text clips were reused from existing output; nine clips were generated for this gallery. No score-based selection, denoising, EQ, pitch correction or gain matching was applied. These are demonstrations, not additional benchmark observations.

Each gallery folder has a manifest with texts, checkpoint and WAV checksums, and provenance. Earlier assets are preserved at their original paths. The family artwork is conceptual, not measurement evidence.

See the preview collection and each model's access and third-party notices. Final model-weight licensing and voice/data distribution review remain pending; this dataset does not imply commercial-use permission.