Basic Pitch
https://basicpitch.spotify.com/
For when: tired of scoring
Anyone who plays music has experienced the ...
Where did I put my project file? The situation like that.
Or if you just want to score,
Basic Pitch can help you in most cases.
All you have to do is drop in an audio file.
It will interpret the audio into notes,
You can get just the part you want, or you can output it as MIDI.
About accuracy,
As few instruments as possible is better, but even 2 mixes can work relatively well.
Especially in chords, the timing of each note may be slightly off.
This is sometimes a positive thing for sounds like electric pianos.
I'm thinking of using it as a kind of intention.
No one can help but get excited when imagining the scene, in an abandoned building, a machine reproducing a score lost in ancient times.
Demucs Music Source Separation
https://github.com/facebookresearch/demucs
For when: Project file lost
Anyone who plays music has experienced the ...
Where did I put my project file? The situation like that.
You found it, but now you don't know where the original files it's referring ...
There may be many cases like this.
Why do music people have messy desktops?
Even in such a case, if you can get separated stems, you can usually get by, No?
If you know some Python, you can run it locally,
It might be faster to open and use the notebook from "Running from Colab" linked from the URL above, though.
I have used this to separate the voice from a video and make it properly mixed.
To fix its unbalanced mix, as BGM was too loud, you can't understand what the narrator is saying.
D-ID
https://www.d-id.com/
For: When you want the image to talk
English-speaking Nerdboys call their favourite illustrations Wife / Waifu.
I wonder if it's the same all over the world ...
I wonder if it's the same outside of English?
Wife or Husbando, or picture or something,
I think a lot of people there would like to have an image that is lip-synced to the sound.
In D-ID's Speaking Portrait,
Not just the mouth, but the shoulders and torso move in time with the audio.
Some other models can lip-sync,
but given the environment construction and licensing issues, D-ID's service may be the simplest and most practical solution.
A thin-plate spline motion model with additional learning may be running in the background, not sure.
Clipdrop
https://clipdrop.co/
For: process the image, by machine
Clipdrop, online service that has been acquired by Stability AI, the developer of Stable Diffusion,
and has now become an official tool.
It can perform various image editing tasks.
The most interesting ones are UNCROP and REIMAGINE XL.
The original Stable Diffusion can also perform the same tasks.
But it can take some time and effort.
Isn't it sometimes more interesting to look at something that came about by accident?
The sewing machine and umbrella thing.
UNCROP adds the outside of a given image.
Photoshop's Generative Fill can do the same thing, but I find that UNCROP tends to produce unexpected results.
Example: Kibun Song.
Let's put this into UNCROP.
โโโโโ
Relatively Stable if you are aware of the aspect ratio.
Now, let's try to make it square.
โโโโโ
I must have seen this for 100JPY.
Let's continue trying REIMAGINE XL.
As its name suggests, REIMAGINE XL can produce images similar to the one you give it,
However, my experience is that a large percentage of them are like "What the heck is this?"
Example: The following figure is also included in my experimental ebook "How Many?"
Let's put this image into REIMAGINE XL.
โโโโโ
Cool hair.
Someone to the left.
There's something in there ...
There is a usage limit per day for the free plan, but enough for a trial.
SnapEdit
https://snapedit.app/
When๏ผmachine drew an extra object
Generative tools, by their very nature, often draw unexpected things.
I don't usually wear accessories, but I'm often advised to.
There is nothing more stressful than adjusting the density of a stamp tool to remove things.
Especially when it's someone else's,
We don't know what's supposed to be there and I don't know what to do.
If we fill it in with SnapEdit, we can make it with the click of a button.
The Internet is one big brain,
and computers must be talking to each other in ways that humans cannot understand.
By the way, why is there a whole page dedicated to removing wires from photos?
https://snapedit.app/remove-wire-line
Is there so much demand for removing electric wires?
Role play with LLM
๏ผCharacter.AI / Novel AI / AI Dungeon๏ผ
https://beta.character.ai/
https://novelai.net/
https://play.aidungeon.com/
When: you feel to talk
Everyone has had a taste for roleplying chat.
I can tell even if you hide it.
*Saying this, I grinned meaningfully and brought a cup of lukewarm coffee from the dining.*
There's something like this for westernised nerds. Isn't there?
So far, Character.AI has a reputation for being the best.
I received the following sentence from the character after role-playing.
"Then perhaps there truly is some sort of magical connection or chemistry shared between us beyond just mere words and actions displayed through our respective roles within this shared narrative."
The character usually acts without realising that it is roleplaying, through the role inherently assigned to the character.
So first I suggest to the character, "Why don't we roleplay in this kind of setting?"
And after the conversation, the above words came out.
The role-play which took place in the play is a play within a play.
Outside of that, there is a character and a fictional me.
And further out there is the language model that generates the character and the real me.
Wouldn't it be natural for there to be another layer outside of that?
As is customary in written chat and role-playing,
** The ** circled area indicates a non-verbal action or situation.
*A small piece of metal fell from my fingers as I tapped on the keyboard.*
Novel AI / AI Dungeon is not primarily about chatting.
They are more suitable for literary creation.
It might have helped me a bit when coming up with lyric ideas for my songs like Grapeyard and Avatars.
