Why I canceled my subscription
I had a Google AI Ultra subscription that I used to make Cinematic videos in Gemini Notebook. Unfortunately, videos continue to have serious problems:
Cinematic videos have a very high-pitch screeching/echo effect in the voice, making it unpleasant and painful to hear. This is the reason why I canceled, as it cannot be fixed by regenerating the video. I have been reporting this issue to Google for months, but they choose to do nothing. These videos are a complete embarrassment for this reason, and I wouldn't share them with anyone.
Explainer videos too increasingly have this pitch problem with a female voice.
Cinematic videos have clips within them that are sometimes cut off prematurely. This is annoying, but it is infrequent, and it can be fixed by regenerating the video, so it is not the reason why I canceled.
Here is an LLM's understanding of the problem:
> The problem with these TTS-generated female voice seems to be more complex than simply having an excessively high average pitch. It sounds unusually shrill, piercing, and fatiguing, with exaggerated emphasis on certain words and particularly on syllables or sounds toward the ends of words. My impression is that this may be a combination of excessive vocal brightness/high-frequency energy, overly large pitch excursions on stressed syllables, harsh resonance or sibilance, and exaggerated prosody that places too much stress, duration, or pitch movement on parts of speech that should sound relatively neutral. The result is a voice that can feel almost screeching at moments and generally unnatural or uncomfortable to listen to. Simply lowering the overall pitch may therefore not fully solve it; the underlying issue may require reducing brightness/harshness, moderating pitch variation and expressiveness, and softening the model's tendency to overemphasize syllables and word endings.
Deep Dive audio within Gemini Notebook, while very good overall, also occasionally has a garbled audio problem. At least it doesn't have the aforesaid pitch problem.
It also doesn't help that the Gemini model has fallen behind dramatically, by multiple generations, relative to other models like by OpenAI, Anthropic, Kimi, etc.
It's like Google doesn't take anything seriously.
My original goal was to create tutorial type videos. Gemini Notebook's Explainer videos can make these for non-technical topics, but not for coding topics where much code needs to be displayed. I now make these using a custom ChatGPT Work Skill in Max Thinking mode, combined with Google Vids, translating a multi-PPTX deck into an MP4 video. I will document my steps in a future post at r/GoogleVidsAI. Examples of created tutorial presentation videos are here and here.