01
Voice identity
Case 3 reports 0.961 mean WavLM-SECS (max 0.975, N=10, one speaker, English, within-WavLM-family). ECAPA-TDNN EXP-1 cross-encoder validation pending. Not a parity claim against any published TTS system.
Chris Royse / leapable.ai
Measured proof - voice case, text case, pending audits
Measured claims stay scoped.
The commercial story is broad, but the measured research claims stay narrow: voice identity, style transfer, embedder counts, and pending audits.
02
Text style
The Shakespeare LoRA case is a public text artifact: bounded corpus, derived labels, model card, and known limits.
03
Pending audits
EXP-1 ECAPA-TDNN validation and EXP-2 pairwise-MI audit remain pending. The realized floor will be published beside the ceiling.
Use the video as evidence context, not a parity claim.
Watch on YouTube· 3:27
Case 3 voice measurement walkthrough
A short walkthrough of the voice-clone artifact and benchmark table. The public claim remains narrower than the video title on YouTube: 0.961 mean WavLM-SECS within-WavLM-family at N=10, one speaker, English; ECAPA-TDNN EXP-1 pending; not a comparison against published TTS systems.
Watch videoOpen the full video on YouTubeHave one context problem?
Send the audience, data type, target task, proof bar, and sharing limits. The first pass should stay bounded.