Thread by @0xCarnival

YOU ASKED FOR THIS👇

💻 Long story short: I run a small YouTube channel about bed stories. I choose books from the public domain, transcribe them and pack them in a nice and interesting way.

Most of these stories are already on Youtube, but they are old videos, which means that most of them have a bad production in all fields (Bad images/footage, bad audio quality, bad packaging).

Now you play with upgraded tools to make those videos look much better:

✅ ElevenLabs for a perfect cristal-clear voice over.
✅ Midjourney or Leonardo for the right images.
✅ LLMs to help you polish your texts.

So let's get into the stuff 👇

Find a site that offers public domain books or audios. I prefer audio recordings, because books are usually in .PDF and the AI has troubble to format the text.

For this example we will use this one: librivox.org/bed-time-stori…

🎧 I selected this audio for the example (right-click into any audio you want) and download it: archive.org/download/bed_t…

Now with that audio downloaded we go to Huggingface and find this cool Whisper Web tool

There, you upload your audio or use the web link and start transcribing.

Some audios like the one I downloaded sounds like crap, so to ensure you get a nice transcription select "quantized" and at least a Medium size model (the bigger the slowest it will transcribe).huggingface.co/spaces/Xenova/…

This is how it looks when it is transcribing. It is slow, but you get the work done, and for FREE.

When it's done, you download the .txt

That file will look like a complete mess. Unless this file is perfectly formatted and the grammar is fixed we can't use this to generate the audio in ElevenLabs, so we will head to openrouter.ai

You can use it for as cheap as $5 to top up some credits.

But "why do you use this and not ChatGPT bro?".

Because for what we want, nothing beats Claude. I've tried with ChatGPT and it's very complicated to make it understand we need a production ready "book" perfectly formatted.

Bear with me, you will see the magic happening soon 👉openrouter.ai

In OpenRouter select Claude 2.1

Remember to set enough "Max tokens" (depending on the lenght of your text).

Then I copy paste the .txt we've got from Whisper and we use this prompt:

I'm working on a bed story book and require your editing expertise to transform this draft into a polished manuscript. While the story has great potential, it would benefit from enhanced clarity, consistent formatting, and occasional sentence reworkings. Please review the text below and help me:

- Correct grammatical errors and typos.
- Ensure consistent formatting throughout the manuscript.
- Revise sentences for clarity and flow, using your creativity where necessary.

[Text from Whisper here]

And let Claude work its magic 🪄

You can see what I mean with "Perfect formatting" pastebin.com/0zZD7dr4

Once you have this the next steps are quite straight forward.

✅ Send the text to ElevenLabs (and please, if you have gone through all this thread and you find it valuable, please, all I ask you is to use or save this voice into your library and use it – it is my voice and I get rewards for every generation you make:

✅ In parallel head to (It is preferred for me, but feel free to use any of your preference) and start creating the images.

I like Leonardo because to my taste it looks better. The prompt I used here is: a girl with blonde long curly hair with poor clothes and poor style, looking up in the sky, talking to the sky, winter at night, snow, in the style of Pixar, 3D cartoon

I am using this kind of Pixar looking images because it's working great to me, and goes well with bed stories.elevenlabs.io/voice-lab/shar…
Leonardo.ai

Then to pack the video I use CapCut, which is free.

It's important that you generate the images and then upload them to the CapCut cloud (via web), that way, everytime you upload your new images you can refresh in your PC version and it will appear there.

👍 When you are done with the video the next steps will be to create a thumbnail and upload it to YouTube.

Will this work? Will you make money? I have no fucking idea. Maybe you make some, maybe you make nothing. It did work for me and it was nice to share with you.

"But why you share it bro? Now everyone will copy your idea!" I don't give a fuck, this is so low effort that it does not make me feel satisfied enough, and I have bigger plans for next year that needs my full focus.

😁 So that was it... again, feel free to try my voice in ElevenLabs and let me know if you liked it and if it helped you bring some more spice to your videos elevenlabs.io/voice-lab/shar…