TencentAuK
AuK
AuK is a 1.5B speech generation and editing model trained on millions of hours of audio. It uses natural-language instructions across several speech tasks.
Model design and access
The release supports reference-based and instruction-based text-to-speech, spoken-content changes, pitch or speed edits, nonverbal expression, enhancement and source separation. A voice can also be described in text without a reference recording.AuK official model documentation
Practical use
Practical guidance. Review the spoken wording and vocal changes independently. When editing one phrase, compare surrounding timing and speaker identity with the original. Use reference voices only with the permission needed for the intended use.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | Downloadable |
| License | MITAuK official model documentation |