Open Source Voice models are here. They are as good as proprietary ones, just no one put in the effort in serving them. We did and it beats deepgram on ALL benchmards. You can try your own.
We made Qwen 10x faster, and 10x more concurrent. Its better in wer in english. Wayy better in multilingual.
We made Qwen 10x faster, and 10x more concurrent. Its better in wer in english. Wayy better in multilingual.