Qwen

/

Qwen3-TTS-12Hz-0.6B-Base

like 225

Follow

Qwen 80.8k

very great work , despite-flashattention (please avoid in the future)

will you be making a combined model ? for all the features in a single model ? ( tts )

also will you make a harness to combine models , such as a omni harness:

so we can choose our asr as a input model as well as a llm and then a tts as a output model ?
also the same for the image pipeline ?
so we can use asr / audio / text as input layer , llm or vlm for processing layer , and tts or image out as output layers ?

these ... conceptual models , will allow for the selecting of custom qwen models as these compoenents int he execution chain , but act as a single model , producing input to output ... via the connected models ... input- processor - output ..

the hugging face had something i was using before as a speech encoder/decoder model .. which is not actually a model but is comprised of two models ? etc ... they had the same for image model etc ..

i think its becomeing a complexed space , with the slim llms they work very well ! as well as the larger models i did not find faults with your online offering to your local models ¬ hence you are the pioneers of this market:

I have also recently - Said to qwen to create podcasts from my sessions, this made the quality of the next generation of the app SUPER ! as it basically used its reasoning from the podcast to generate the next iteration !

so :
will you implment the ability for the model to generate an output with multiple voices ? so if we request to make a podcast with characters it will select the speakers accordingly or you maybe can specify speakers ?

also the qwens have become very great in the create a discussion arena !!!