Repository navigation
wllama 3.13.0: Welcome new ASR π¬, TTS π£οΈ and realtime πββοΈ capabilities!
This release added support for ASR, TTS and realtime support, which allows interacting with realtime speech-to-speech models like the LFM3-Onmi series.
Try it here --> https://github.ngxson.com/wllama/examples/omni/
What's Changed
- [autosync] llama.cpp 2bc5635 by @ngxson in #290
- [autosync] llama.cpp 5c8854f by @ngxson in #292
- [autosync] llama.cpp 5aeac37 by @ngxson in #293
- Add
originsoption tocrossOriginStoragerequest and rename togetFileHandle()by @tomayac in #262 - scripts: guard the missing wasm and the uninitialized submodule by @ServeurpersoCom in #282
- Image embeddings, multi-embeddings, embd_normalize by @thomas-0816 in #295
- sync latest llama.cpp source code by @ngxson in #303
- put source map to wasm by @ngxson in #304
- add ASR and TTS (+ realtime) by @ngxson in #313
- add omni example by @ngxson in #317
- Treat a model with a missing shard as invalid instead of throwing by @felladrin in #315
- npm stage publish by @ngxson in #318
New Contributors
- @ServeurpersoCom made their first contribution in #282
Full Changelog: 3.8.1...3.13.0