Overall, a pretty solid app. I’m kinda new to voice cloning, so take this with a grain of salt. The controls are simple and fairly intuitive. The cloned voice is clear and annunciates, but I wouldn’t say it matches the original voice. I tried a few different samples-none 100% matched the original speaker, most only produces a slight resemblance to the original. For my application, it is more important to produce a clear, easy to understand voice than to be able to recognize/identify a specific speaker’s voice. I would also note that it does speed up the output voice. In order to produce a steady output, I had to drastically slow down the input-borderline obnoxiously slow. For features/improvement: I couldn’t find a way to store voice samples on the app. Also, the adjustments for pre and post generation seem different. I’d like the same sliding scale/options on the input side, then be able to go in and fine tune an individual segment if needed. Finally, a way to easily adjust the tempo or speed without affecting the voice pitch would be nice. Again, fairly new to voice cloning, but for producing somewhat natural sounding voiceovers - especially for not requiring a subscription and keeping everything local - it’s the best option that I’ve found.

This is a lot of fun and works fairly well on an iPhone 15 Pro Max. Short text inputs all worked without issue, but as soon as the text got longer, crashing was much more frequent. Not sure if it’s still hitting OOM or what. I love that this is offline and doesn’t collect data and I understand that comes with limitations. One recommendation would be to add a way to support multiple TTS models but just default to your favorite for simplicity. That would help it scale as more models come out with different tradeoffs in performance. Looking forward to where you take this.

I needed to replace a sentence in a recording that the narrator flubbed. I downloaded Chinny, turned off WiFi (to confirm it wasn’t processing anything in the cloud), loaded a 30-second sample of the narrator’s voice, typed the script for what needed to be narrated, and got an amazing result. While it didn’t exactly match the previous recording (it sounded like the clip was recorded on another occasion by the person), it certainly sounded like the narrator, down to inserting realistic inhale sounds between phrases. The cadence was too fast (which I fixed easily with other audio-editing software), but overall sounded very natural and matched the style of the narrator. And it took me maybe 10 minutes start to finish, including downloading Chinny in the first place. I’m very impressed.

It’s a great little tool and proven useful for projects now. However it’s basic with a lot of room for improvement. There are no emotional tags support in text prompts for finer deliver; no speed or pitch controls, but there are a few options. Very hard to predict or control the outcome of elongated words and punctuation to achieve more naturally speaking audio.Hoping for updates!