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](https://huggingface.co/collections/owensong/inflect-v3-preview-6adad1c0c9e6fb018f440610) 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.