

In the purchase for the perfect bronze paint for a white car. What do you think of this one?
The car is a white 2021 Mazda 3. I had already one iteration where I failed a little with choosing the right tone and it ended up being more of a satin gold. Now I'm about to try again. Should I go with this one?
I mean.. what do you even play after this?
I'm definitely going for the second round, but right now I'm just exhausted after finishing it, so will have to find something else.
I’m more of an outsider to this topic, not per se a TTS specialist
It’s weird to me that text normalization still feels so underdiscussed in streaming TTS.
I see a lot of talking about latency, naturalness, voice quality, expressive speech
but models surprisingly start looking weak on basic everyday stuff like prices, dates, phone numbers, and all the usual letter-number mess. Started noticing a lot in cars systems
Maybe I’m missing something, but most benchmarks I’ve seen seem way more focused on how nice the voice sounds than on how the system handles messy real-world input in a streaming setup
So for people deeper in voice / TTS:
is this just a normal unsolved pain point everyone works around or it’s just the case witn in-car assistants?
do solutions already exist?