It is worth mentioning that most of the Eleven Labs characters will go into getting the reading just right... it is best sounding but also the hardest to control.
Regeneration does use up characters. It irritated me at first because I went in expecting 100 characters used would equal 100 characters of finished product. I've adjusted my perspective a bit by comparing it to filming and editing. Maybe it's just me, but when I record a video, I usually have multiple takes and lots of footage that aren't good enough to use. 60 mins of recording might get me a 15 minute video. Even though I'm not using "credits" when I'm recording myself, I'm using time ... Recording takes ill never use and then editing them out. Credits cost a little money, but getting to a finished product in elevenlabs is much faster (in my experience) than recording and editing myself. That's just how I see it, but everyone else's experience might be different.
@@excelerator Yeah, I guess it makes sense when 11L does what it is supposed to just doesn't do it well... but when the audio gets quiet with every sentence, or it spits out sentences at 100mph it get's really annoying...
If you're going to do testing, it's best to have the AI try to recreate the same script that was used to train it. Play it back to back and compare it to the real voice.
Fair point. I guess if you have something you've already recorded and used in a project, you wouldn't need to record a dedicated sample. If you don't, it might be worth it to make the sample recording in exactly the style and tone you want so that you can use the clone from that point forward and "never have to record again." Depends on the individual situation. Might be worth it for some and absolutely silly for others.
In that test, I was really impressed with PlayHT. My go to is elevenlabs... I've made some adjustments and getting better results now (I agree, it errs on the faster side.) I'm watching descript to because their AI voices are getting better.
Play HT 👉 excelerator.tech/GetPlayht
ElevenLabs 👉 excelerator.tech/getelevenlabs
Thanks for the video. I enjoyed the analysis. Great pacing and result.
Glad you like it!
Your voice is one of a kind, no AI can understand it yet.
I'm glad my voice is one of a kind... At least I don't have to hear it that much. 🙉
It is worth mentioning that most of the Eleven Labs characters will go into getting the reading just right... it is best sounding but also the hardest to control.
Regeneration does use up characters. It irritated me at first because I went in expecting 100 characters used would equal 100 characters of finished product. I've adjusted my perspective a bit by comparing it to filming and editing. Maybe it's just me, but when I record a video, I usually have multiple takes and lots of footage that aren't good enough to use. 60 mins of recording might get me a 15 minute video. Even though I'm not using "credits" when I'm recording myself, I'm using time ... Recording takes ill never use and then editing them out. Credits cost a little money, but getting to a finished product in elevenlabs is much faster (in my experience) than recording and editing myself. That's just how I see it, but everyone else's experience might be different.
@@excelerator Yeah, I guess it makes sense when 11L does what it is supposed to just doesn't do it well... but when the audio gets quiet with every sentence, or it spits out sentences at 100mph it get's really annoying...
@BetterTimers Absolutely! I'm not saying I love getting charged for the bad takes... I just figured out how to not be AS frustrated about it.
If you're going to do testing, it's best to have the AI try to recreate the same script that was used to train it. Play it back to back and compare it to the real voice.
Excellent point! Thank you. I never thought about having it generate what I trained it with.
If you can upload two to three hours of sample audio, do you really need a voice cloner?
Fair point. I guess if you have something you've already recorded and used in a project, you wouldn't need to record a dedicated sample. If you don't, it might be worth it to make the sample recording in exactly the style and tone you want so that you can use the clone from that point forward and "never have to record again." Depends on the individual situation. Might be worth it for some and absolutely silly for others.
Play HT produced the best output, IMO. ElevenLabs was too fast. Like a speed reading contest. Which one is your go to platform?
In that test, I was really impressed with PlayHT. My go to is elevenlabs... I've made some adjustments and getting better results now (I agree, it errs on the faster side.) I'm watching descript to because their AI voices are getting better.
I find Play HT better than Eleven Labs in every way. It's just a shame they don't have a cheaper plan like Eleven Labs does.
There's a lot to love about PlayHT. 👍🏽