How to make the voice of FFW
(and any other voice)
Here’s an instruction on how to clone the voice of FFW or any other voice you
can imagine.
I will explain the following process to you step by step:
I. Search for YouTube videos
Search for YouTube videos that contain the voice you want to clone.
THESE YouTube video must meet the following conditions:
1. Only contain the voice you want to clone,
no other
voices
2. Be shorter than 10 minutes
3. No sound effects such as delay, no
harmonies
Copy the links of the YouTube videos and save them for later.
II. Clean the audio
If you want to clone any voice, you need to create your own sound model using
AI.
The following principle always applies:
“In terms of quality, you get out, what you put in.” https://www.youtube.com/watch?v=TV-NVRKtHVo
So if the YouTube videos contain any form of background sounds or music, you
need to clean them first.
1. Go on “Kits.ai” and “Vocal remover”
2. Enter the YouTube URL
3. Enable “Remove Instrumental”, “Remove
Backing
Vocals” and “Remove Reverb”
4. Go on “Split Vocals”
Now there will be 4 different files. You only need the first of these called
“Vocals”.
III. Train your voice
Now it’s time for training your voice model. To get optimal results, we need 30
minutes of cleaned audio.
Go on “Voices” – “Train a voice” – “Continue” and then upload your 30
minutes of cleaned audio.
Select “Speaking” and name the model “Far From Weak” for clarity. You can
also add a picture if you want.
Then you need to wait for the training process to finish. This might take a few
hours.
IV. Getting the input
To get the input, we need to go on “use voice” and then upload the audio file.
So basically we have a Text-to-speech model here, not a speech-to-speech
model. In the following I will briefly explain what you need to know about it:
In the first variant, a text is used as input; in the second variant, you have to
record your own voice and feed it into the AI language model.
However, both methods have their downside:
If only text is used as input, the voice sounds quite monotonous because you
cannot create the desired accents in the flow of speech. For example, you can
not create the “pathetic” of FarFromWeak with this method.
On the other hand, if you use your own voice, the result often doesn't sound
nice either. If you have bad pronunciation, the mistakes will be picked up by
the AI. The same thing happens with a poor intonation of your own voice. In
short, this method only makes sense if you have a really good voice.
But there’s a solution for this problem: We just combine the two methods!
Go to Elevenlabs and sign in with an Email address for free. Select the voice
called “Brian” and type in the text you want to generate.
Choose the model “Eleven Multilangual v2” for getting the best quality
possible.
Now you can download the audio file. You have 5.000 character per Email-
emphasise on the “per Email” part. You get what I mean.
Let’s bring the file back to Kits.ai and listen together to the results:
Put the pitch level at -2, conversion strength and volume blend both at 0% and
then press “convert”
As you can hear, the result doesn’t sound exactly like FFW. In the next step,
we’ll fix this issue.
V. Improving the quality
Here you can see the improved process diagram to get a higher quality.
Go to „weights.gg“ and search for “Kratos”. Select the one called “Kratos God
of War Christopher Judge” and download it. Upload the file to “Kits.ai” and
name it “Kratos” for clarity.
Now bring the audio file to the “Kratos” model and use the settings from
above.
Here you have Kratos' voice, which is much more suitable as input than
Elevenslab's original voice.
To get the best possible quality, clean the voice as mentioned above.
It's finally time to get the result! Bring the file back to the trained FFW model
and put the pitch level at -4 this time.
This time, the output sound exactly like FFW.
Chapter 2: Danger alert
As you may have noticed, the video is not over yet.
Daher wird YouTube gegen geklonte Stimmen in Zukunft umso härter vorgehen.
https://www.youtube.com/watch?v=G8EdGyytN_s
I have bought ChatGPT 4 to get better results. If you send me your main niche, I
can generate those for you so that you can achieve the best possible results.
VI.Avoiding copyright strikes
You can’t use these voices because there’s the danger of copyright strikes. Even
if it is not the voice of a celebrity, other creators on YouTube might also use the
same model.
The algorithm can now use various metrics such as demand, competition and
average view duration (=AVD) to determine which niche has the highest
potential. This requires the cheapest plan ($12).
Now simply copy the first suggestion from ChatGPT into the search bar. Press
"Search" and then go to "Related Keywords."
Then copy the first line into an Excel table. Now do this with all 50 keywords.
VII. Evaluate Excel table
After you have finished, your Exel document should look like this:
Now copy the spreadsheet for the case you need the information later.
Next, delete the second and last column for clarity.
Then mark the last column and go on “sort & filter” – “Sort largest to smallest”
And now you have the final result! All listed sub-niches with a score of 60 and
above are very good sub-niches: There is very little competition in these sub-
niches and the topics are in high demand at the same time.
VIII. Outlook into the future
YouTube will tighten its policies regarding AI-generated content, including AI
voices. The platform has introduced new rules requiring creators to disclose
when they use AI to generate content.
So it will be seen in the future whether an AI voice is enough for a warning to
be displayed on your own videos.
https://odishatv.in/news/technology/crackdown-on-ai-generated-videos-on-
youtube-soon-220225
https://www.youtube.com/watch?v=M09F_fLqHAc#:~:text=URL%3A%20https
%3A%2F%2Fwww,100
To use this feature, you need a paid subscription.
Like last time, I will share my account with the first person who asks me.
If you were too late, there is another option:
Everyone who works with me on the YouTube project will receive access from
me.
👉 Please write me your email address if you want to take part!
This community project is about us creating YouTube videos together to learn
from each other.
We have already agreed on a subniche, a channel name and a channel logo.
The goal is to upload the first video by the end of the month.
The concept of the project only makes sense if we work on it together and
exchange knowledge.
