More Language Support
Great work on the model and GGUF support for llama.cpp.
Adding support for more languages would make this model significantly more useful for local, offline audio processing, especially on non-English systems, mobile devices, and global use cases where internet access or cloud APIs are not available.
This would greatly improve accessibility and real-world adoption.
Absolutely. We are in progress of supporting all languages supported by LFM2.5-Base also within LFM2.5-Audio. This future support will cover both multilingual input as well as output (generation).
Also we could use voice design by prompt 😄 just like Qwen-TTS 👍
Future plan, never mind.
If you'd provide a small readme / notebook on fine-tuning with a specific language dataset, the community could assist.
Hello,
at:
https://www.liquid.ai/blog/lfm2-audio-an-end-to-end-audio-foundation-model
you say:
Deploy and build with LFM2-Audio
...
Emotion detection
How can we ask to the model to perform the emotion detection during ASR?
Thank you so much for the great work. Best!
Hi, emotion detection is not supported in ASR mode in this checkpoint, it would have to be specifically finetuned unlock this capability. Public finetuning support is coming though!