mastodon - Link to source
The amount of times Google Gemini gave me a solid code correction and answer while Chat GPT sits there thinking and thinking is staggering, honestly it's making me disavow it because it's not like this is better in "pro" mode. I noticed Gemini "thinks" in smaller chunks but it could be just that they update the process more often. Regardless, GPT, you're beginning to really really suck. Gemini can give me the same quality of C++ and Python code recommendations, it's still shit when pasting back full code for you so better when patching, which is about the only thing holding me back from it as opposed to GPT. Gemini can read attachments but it doesn't quite get the same containerization treatment - GPT mounts your files from /mnt/data, Gemini clearly stores it somewhere and can read but when writing it is restricted from spitting back a file for you and envoking commands in a container. Huge downside to Google, ugh.
Something that using a local Gemini Cli instance could solve though, honestly.
This entry was edited (7 minutes ago)

mastodon - Link to source

I bloggered a post.

It's about shortcomings of FLOSS and a possible next thing.

My Next Project Won't be FLOSS:
pointless.one/my-next-project-…

#FLOSS #FOSS #FreeSoftware #OpenSource #GNU #GPL #OSI #MIT #BSD #BTPL #PolyForm


My Next Project Won't be FLOSS @ PointlessOne

Where FLOSS fails and what to do about it. With a little bit of history.
PointlessOne
#FreeSoftware #foss #opensource #floss #bsd #gnu #gpl #OSI #mit #btpl #polyform

mastodon - Link to source

Some of the variables you can define in the yaml now:
# Base defaults for NV Speech Player frontend.
#
# If a setting is not listed here, the C++ defaults are used.
# Packs are merged in this order:
# default.yaml -> <lang>.yaml -> <lang-region>.yaml -> <lang-region-variant>.yaml
#
# Keep this file small. Put language-specific tweaks into their own files.

settings:
# Timing / stress shaping
primaryStressDiv: 1.4
secondaryStressDiv: 1.1

# Stop closure insertion ("click" before stops/affricates)
stopClosureMode: vowel-and-cluster
stopClosureClusterGapsEnabled: true

# Length mark (ː) handling
lengthenedScale: 1.05
lengthenedScaleHu: 1.3
applyLengthenedScaleToVowelsOnly: true

# Output gain defaults (same as ipa.py)
defaultPreFormantGain: 1.0
defaultOutputGain: 1.5

# Normalization cleanup
stripAllophoneDigits: true
stripHyphen: true

# Tonal languages
tonal: false
toneDigitsEnabled: true
toneContoursMode: absolute

This entry was edited (26 minutes ago)

mastodon - Link to source
Sigh. We could pass the cast pointer from the callback since it points to valid C++ memory, but it could be riskier in some cases, especially with temporary DLL memory. The safer approach would be to create a new speechPlayer.Frame, set its fields, apply modifications, and then send it to queueFrame. We also need to handle cases where the frame is NULL, meaning silence. So much work.

mastodon - Link to source

So I wasn't going to post about this but I just got off the phone with Tesco customer service and I'm shocked.

Last night I was having some soup, Tesco brand Tuscan inspired Bean soup, to be exact, bit down on a chunk of glass that was in there. Taste of blood in my mouth, dent in my tongue, possibly swallowed a small bit... I was worried about it. Also, fucking hell like.

Today I call em up, had already sent an email, they are uninterested, tell me I have to complain in store.

What the fuck.

⇧