Ashley

co-founder @liva_ai (yc s25) | prev cs @caltech | sometimes i dance

[Ashley @ashlleymo]

@ycombinator @_eosai @ary_src Luv it!!

[Ashley @ashlleymo]

@morganlinton thx morgan!

[Ashley @ashlleymo]

fought so bad for a one way tix to sf. proud of my fellow internationals who also made it here ❤️

[Ashley @ashlleymo]

@AmolParikh40001 moved!

[Ashley @ashlleymo]

@emreshepherd No good photos today, so took one from summer

[Ashley @ashlleymo]

Lmk if anyone has recs to meet new ppl in the city (e.g. play sports, dinners, etc.)

[Ashley @ashlleymo]

@avikonduru joining y’all in the land

[Ashley @ashlleymo]

love this new website, and will always love this community

[Ashley @ashlleymo]

https://t.co/JXfQlqG38f

[Ashley @ashlleymo]

Qwen3-TTS Family is now Open Sourced!

- Supports 10 languages
- Has voice cloning and design abilities
- 5 models (0.6B and 1.7B)
- High compression (12Hz)

Highlights:
- Qwen3‑TTS‑VoiceDesign beats MiniMax‑Voice‑Design and other open‑source models on InstructTTS‑Eval for instruction following and expressive generation
- Qwen3‑TTS‑Instruct achieves an average multilingual WER of 2.34%.
- Qwen3‑TTS‑VoiceClone achieves 1.835% WER and 0.789 speaker similarity on the MiniMax TTS multilingual test set across 10 languages, outperforming ElevenLabs and MiniMax

Congratz to the team!

[aoi @bigvalleyblue]

NVIDIA's Personaplex model is super impressive, enabling natural, human-like conversations while responding with nods and affirmations!

Super natural sounding + can run locally at 7B size.

- Inspired by Moshi from @kyutai_labs
- Full duplex, which means AI listens while talking (no more awkward AI pauses
- Handles interruptions, backchannels, turn-taking

Congrats on the team!!

[Ashley @ashlleymo]

https://t.co/NXvJnZFoAb

[Ashley @ashlleymo]

Just finished reading a paper showing that modular speech-to-speech systems can achieve sub-sec latency like end-end models while having fine-grained controllability.

Summary in thread 🧵

[Ashley @ashlleymo]

Future work (6/6)

- They plan to introduce models like acoustic echo cancellation to improve stability

- May also add more generative models to get more flexible and expressive audio output

- Congrats to the MoE Key Lab of AI and X-LANCE Lab (SJTU), Shanghai AI Laboratory, BIGAI, Fudan University, Northwestern Polytechnical University, and SIAT-CAS for this paper