I generated 100 videos to get viral on TikTok using Google Cloud free credits and it wasn’t worth it.

I warmed up the account over a few days and then started posting videos.

I started by just posting glass fruit cutting videos, since those sounded the most promising, with hashtags as description. Did fine, but not too many likes.
The first video, a glass banana, was posted on 7th of October and got 4 likes and 208 views total.

Off to a weak start ^^

I then posted one video each day for the next few months.
While writing this we are currently at ~70/100 videos.

The goal was to reach at least 100 followers and get at least one viral video because getting monetized isn’t even possible in Austria since the TikTok Creator Rewards Program is only available in select countries like the United States, United Kingdom, Germany, Japan, South Korea, France, Mexico, and Brazil. Austria has never been included and isn’t expected to be anytime soon and I knew this before starting.

I failed and barely got past 1k views per video, ending up with 61 followers and around 952 likes at 70% completion of the challenge.

The glassfruit idea went ok, but a lot of accounts posted similar content with worse quality while getting more views, likes, and followers. Some might have been bought accounts, but that’s just speculation.

I generated the first batch of glass-themed videos over 2 days and then scheduled them.

It is a bit annoying that TikTok only lets you schedule for the next month though. That meant I couldn’t just generate all videos and leave the account unattended for the next 3+ months and check results after.

I posted them at different times to find out the best time to post. 4 p.m. was generally the best.

After like 25 glass-themed videos I started experimenting a bit more.

I made some with slime, sand and soap. The slime ones worked out a bit better.
I generated some hydraulic press videos that went ok but didn’t turn out that well.
I tried only posting those that had relatively proper physics, but I didn’t have enough videos in stock at times so I posted some slop too, sadly.

I continued with more realistic ones like slime balls and mixed in a realistic minecraft block which went a lot worse than I expected and only got 10 likes and 373 views as of now.

By the way, the first few ideas were researched using AI Studio with Gemini 2.5 Pro and I asked it to search for the most popular fruits on Google since I thought those would work the best. Yeah, it was okay haha. Continued with some more tropical ones and mixed in a few minerals from time to time.

I started getting more creative on 10th of December and made some more abstract art and more interesting videos from now on.
Those went a lot better.

I could have just made more similar ones of the best performing ones, but that seemed too easy and the sample size isn’t even that big for less than 100 videos so it could have just been luck.

By the way: if you paste in hashtags without actually clicking at them and selecting the one you pasted it doesn’t paste them as hashtags but text, which is really bad UX in my opinion but what do I know?

I started adding a short description to the hashtags on 15th of November which might have helped somewhat. I am not sure.
I consistently didn’t get too many views, likes, bookmarks or comments. The most I got was from other accounts doing a similar thing…

I labeled all of them with the AI label if I didn’t miss any. This might have also limited reach, I am not sure.

One video got taken down for some reason but was added back 30 minutes after I disputed it.

Statistics

Top Posts

Viewer Analytics

Audience Demographics

Best Posting Times

Coolest results in my opinion

How did I generate the videos?

I used Google Cloud Vertex AI to generate the videos with the ~$250 credits you get when you sign up. I also used a few more credits from a friend who also signed up around that time and said he wasn’t using all of them anyway. So all I paid was time and patience: Generating was free, although I almost paid a bit since Google always lags behind many hours till the UI finally updates. I lost track of spending at the beginning since I was using the UI in Google Cloud itself which is hidden somewhere in this horrible UX mess named Google Cloud Console.

I made a simple UI if you want to try the same challenge as me and open sourced it here. Demo at veo.merginit.com.

Why did I do this? I noticed that after every generation the UI state resets to the initial state, causing me to accidentally generate videos with invalid settings a few times which caused me to spend a lot more and lose track of spending.

This UI keeps a history in your local browser IndexedDB, so that you don’t accidentally lose generated videos and prompts on refresh or connection loss. I also included a simple prompt generator in it.

You can technically store generations in Google Cloud Buckets too, but that costs you some money.

Failed generations - Top 15

Error Codes I encountered

I didn’t spend too much time on making the perfect prompts and mostly used AI to generate them from my ideas based on some AI research and other videos I watched on TikTok that went viral in the same genre.
About 25% of the generated videos didn’t reach my expectations.

Weird error code I got two times that didn’t make much sense, since it only failed for those 2 prompts:

Vertex AI Error (401): {
  "error": {
    "code": 401,
    "message": "Request had invalid authentication credentials. Expected OAuth 2 access token, login cookie or other valid authentication credential. See https://developers.google.com/identity/sign-in/web/devconsole-project.",
    "status": "UNAUTHENTICATED",
    "details": [
      {
        "@type": "type.googleapis.com/google.rpc.ErrorInfo",
        "reason": "ACCESS_TOKEN_EXPIRED",
        "domain": "googleapis.com",
        "metadata": {
          "method": "google.cloud.aiplatform.v1.PredictionService.FetchPredictOperation",
          "service": "aiplatform.googleapis.com"
        }
      }
    ]
  }
}

Vertex AI safety filters blocked my prompts about 10% of the time:

Vertex AI safety filters blocked the video: Veo could not generate 1 videos based on the prompt provided. You will not be charged for this request. Try rephrasing the prompt. If you think this was an error, send feedback. Support codes: 00000000

The cheapest way

Recommendation: get a Google free trial so that you can test in there for free first (3 per day) and then choose the prompts that worked best there.

Takes a lot longer but is cheaper and produces better results.

You can test ~30 times 3 = 90 prompts that way for free, maybe even ~30 times 5 if they increase the rate-limit again.

Why this was stupid

I think YouTube would be a better place to post them since ad revenue is a lot better there and I could actually get monetized there.
I didn’t post there though and wanted to test the TikTok waters, since the only substantial account I have ever built was a meme account on Instagram. It might be interesting to also post the videos on YouTube in the future and compare the results, but I don’t think it would really be worth it.

The TikTok account

Check out the experiment: @aiasmrlvr on TikTok