Lon Ligrin

@AsgardAGI All work. Buuut you might need a good reference audio for each one 🤘🤠

@LonLigrin OmniVoice ? do you have a list of which emotions work?

Made another emotions quality improvement. Also added some more emotions: disgusted, amused, relieved, embarassed, disgusted

Sneak peek at RealtimeTTS 0.6.1 👀 🔥 Omnivoice support 😡 Inline emotion tags ( [anger] ) ⏸️ Inline pauses ( [pause=0.8] ) Test it on GitHub today: github.com/KoljaB/Realtim… #TTS #Omnivoice

@sparbuchfeinde 3.) Wir sind technologieavers und überreguliert. Profite durch neue KI Firmen werden woanders entstehen.

Deutschland wird einer der größten Verlierer der KI Entwicklung sein. 1.) KI braucht Unmengen an günstiger Energie. Deutschland hat die höchsten Strompreise der Welt. 2.) KI macht Mitarbeiter obsolet. Amerikanische und asiatische Firmen können kurzfristig zehntausende Mitarbeiter freisetzen. Deutsche Firmen müssen sich mit dem hiesigen Arbeitsrecht herumschlagen. Mitarbeiter sind praktisch unkündbar und nur über sehr teure Abfindungsregelungen abzubauen. Beide Faktoren sind massive Wettbewerbsnachteile.

It works so well. Hm.. give it a voice interface and make it more universal? Like not only installing github repos, more allround CLI assistant?

Code: github.com/KoljaB/github-…

KoljaB/github-install-assistant installing hermes agent in wsl2 for Windows youtu.be/anxVHepsxEA

Disclaimer: performs LLM crafted shell commands on the CLI. Use responsibly, there is always the potential that tools like this can do damage to your PC.

python github_install_assistant.py github.com/KoljaB/Realtim… --goal "Install RealtimeTTS with faster qwen3. Create a new venv. Install PyTorch (cu128) first via the correct CUDA index URL." --target-dir D:\Test --planner-model gpt-5-nano --loop-model gpt-5-nano --max-steps 40

Wanna try that shiny new repo? But last time you met CUDA hell? Here's a repo that installs other GitHub repos for you: github.com/KoljaB/github-… It does not import anything, so it's easy to install.

You can also do "pip install RealtimeTTS[qwen]==0.6.0" now. Check the repository nevertheless or you might miss out for example the code for recreating the videos. For example tests/faster_qwen_emotions.py from gh repo recreates this: youtu.be/ZOKcUpJlrXQ

Generating emotional AI voices in real-time. youtu.be/ZOKcUpJlrXQ?si… Tech stack: - faster #Qwen3TTS - #RealtimeTTS - Ears Dataset by Meta.

Looks great

Check out faster qwen (or one of the other 6 new engines) in latest RealtimeTTS. (pip install coming soon)

RealtimeTTS 0.6.0 new on gh With 7 new engines: Qwen3 TTS, Cartesia, MiniMax, ModelsLab, CAMB AI MARS, NeuTTS, PocketTTS Real-time multi-user streaming (WebSocket) Volume control 16 PRs merged + lots of improvements pip install coming soon

@waltertayannlee Using two streams doubles compute and prob puts you into merging hell. English model will try to transcribe Mandarin and vice versa. You'll be picking higher-confidence results and juggling with timestamps. Would only try if inputs are separated, like one channel per language.

@LonLigrin if you had to transcribe both Mandarin and English what would you use? I'm trying Deepgram but their multimodal options are limited to a few languages. I'm considering just opening 2 websocket streams (one for English and the other for Mandarin) then just combining the output. I've tried the Deepgram discord support and this is what the bot seems to be suggesting? 😭😭 Do you think this is best practice? Or is there something I'm missing. Sorry I would DM you but your DMs are closed!

@waltertayannlee Changed dm settings. With opensource tools only I don't see a really reliable solution for real-time multilingual transcription. So from what I heard I'd prob go with Soniox v4 Real-Time. But take that with a grain of salt please, since I don't have any rl experience with it.

@Yuchenj_UW It's a cliff not a gap. Companies won't hire the best human, they’ll just hire the one human they can't automate.

Stop worrying about AI replacing you. Worry about this instead: Are you + AI > others + AI? That gap is your moat.