Affordable text to speech AI alternative to ElevenLabs

[Rauru]

Hi there.

I'm starting a YouTube channel this way:
1. Downloading a youtube video of the topic you wanna publish in let's say French.
2. Converting audio into text with ElevenLabs or similar.
3. Translating into let's say English.
4. Using some AI like ElevenLabs to convert the English text into English audio.
5. Making a video and uploading it to YouTube.

The problem comes if you wanna upload something like 1 hour long video weekly; steps 2 and 4 are nightmarish bottlenecks.
ElevenLabs spends many credits already on step 2.

But the real problem is step 4 text to English voice. Here the ElevenLabs subscriptions will bankrupt the heck out of you.

So, is there anything similar to ElevenLabs for a reasonable price allowing to generate like 4h text to speech per month?

 

[CloudMerchant]

for step 4 you dont need elevenlabs honestly. that pricing is brutal once you go past the free tier. couple options that work fine for long form:

- kokoro tts is open source and runs local, basically free if you got a decent gpu or even cpu for short batches. quality is surprisingly good now
- fish audio / fish speech, also self hostable
- if you dont wanna mess with local stuff, look at deepgram or playht, their per character pricing is way cheaper for bulk

for 4h a month local is the move imo. the upfront pain of setting it up pays off fast vs paying eleven every month.

also for step 2 (transcription) stop wasting eleven credits on that... use whisper. its free, runs local, and faster-whisper does an hour of audio in a few min. theres no reason to pay for STT at all when whisper exists.

one thing tho, if youre reuploading other peoples videos translated, watch out for the reused content / copyright stuff on yt. monetization can get rejected if its too close to the source. adding your own commentary or restructuring helps. just a heads up.

 

[Fluxify]

Whisper + kokoro combo is the way like Cloud said, been running that exact stack for a few months now. One thing nobody mentioned, kokoro can sound a little flat on a full hour of narration... if your channel is more documentary style its fine but for anything emotional it falls apart. What i ended up doing is kokoro for the bulk and saving eleven credits only for the intro/hook part where the voice actually matters for retention.

Also faster-whisper large-v3 is worth the extra vram over the base models, the timestamps come out way cleaner which saves you headache on step 5 when youre syncing.

The copyright thing @CloudMerchant flagged is real btw, seen channels get smacked even with translation. Translation alone isnt enough transformation in yt's eyes anymore. Throw a short intro, cut dead air, maybe overlay some b-roll and youre usually fine. The pure reupload-but-translated play used to work great like 2 years ago, much harder now imo.

 

[flackz]

Install Microsoft Clipchamp, and use it for free for all your text-to-voice needs.

 

[XSMMDoctor]

Piper is worth a look too if you don't care about the voice sounding “premium”. I used it for faceless recap stuff and it was good enough once mixed with a bit of music/room tone, but raw voice for 60 mins can feel robotic.

For 4h/month I’d probably do Whisper locally for transcript, then either Kokoro/Piper local for bulk, and only use ElevenLabs for the first 30-60 sec hook like @Fluxify said. That part matters more than the rest anyway. Clipchamp works but I wouldn’t build the whole workflow around it, feels clunky when you’re doing long scripts every week.

 

[r2150523]

Sorry but if your first step is to download someone's video, you're stealing content and YouTube filters this very quickly these days.

You'll get banned or your videos won't rank as well as new ones would.

It's a bad method.

If you can't produce content, outsource it.

Yes, it's a nightmare that black hat is dying. But that's how AI works. You can't beat it. It's a waste of time you could spend on providing real, fresh value to the system.

Think of real business skills instead.

 

[r2150523]

one thing tho, if youre reuploading other peoples videos translated, watch out for the reused content / copyright stuff on yt. monetization can get rejected if its too close to the source. adding your own commentary or restructuring helps. just a heads up.

Exactly this. Do something like automated v-tuber content instead. But automation of this is expensive. Building software just to do that is like starting a business. And your channels still will risk being banned.

 

[Rauru]

Many of you mentioned I shouldn't download content from another people.

But what if I download a video feom YouTube, convert it to text, then change the text with AI and add more content (or remove some part to add something original), after that I use text to audio?

Would that work?

Also, what if I use my own voice? I'm not a native English speaker but the voice would be original (maybe similar to another guy but original). Would the audience accept it?

 

[flackz]

But what if I download a video feom YouTube, convert it to text, then change the text with AI and add more content (or remove some part to add something original), after that I use text to audio?

Would that work?

Yes, of course. This is a common method.