The AI Content Creation - 2025 state
Premise:
Can an individual in 2025, with no background in content-generation, use an AI platform to reliably produce content and profit from it?
The above was the question that the following idea was based on.
The idea
The original idea was to automate the entire video creation flow for the content creators. So, basically, the user will enter his search topic, and the search proceeds in the usual deep search way(Langchain's open sourced Deep Search) but the final report generation is modified to produce veo-3 prompts with content for voice over for that video.
Realization
As I progressed through the work, I realized that it was more complex than I thought. Because, the videos were generated using Veo-3 but how will you give the voice over. For this, again you need to a tts(test-to-speech) model. And then you need to piece the video and the speech in a coherent way which involves editing(again a lot of manual work).
I generated a 30-secs video with a random topic, just to understand the whole process. I asked the research agent to come up with an interesting topic that was viral in 2005. It came up with some Chuck Norris facts and generated some 6 prompts along with the voice over content(The actual core content). Then I proceeded to generated the videos. Once the videos were available, then I proceed to use 'Eleven Labs' studio to generate the voice over and piece it with the video. The video was a slop but I understood the process involved. (Moreover, I spent 2000 rupees on this AI slop(only from Veo-3) and this is excluding the free credits I used from Eleven labs which if I add, would go up to 3000 to 3500)
Only after this I realized that automating the content generation entirely with AI is really complex. The best one can do is, when thinking about creating a content creation platform, is to put up a platform like Eleven labs studio, where you basically give the user, the traditional editing tools along with embedded AI tools to generate video, images, text-to-speech, voice over and other things. Because there are different styles of video creation like Voice Over(Face less), In person video, mixed, podcast styles and then there are adds and so on...
Takeaways:
The video creation using AI is still a costly affair. At least in 2025. Not sure about the progress of Open Source models but it is definitely a costly affair in closed source models that too via APIs(veo-3, Sora).
- Relevant links: google_veo3_pricing_almost_killed_my_ai (Seems the trick is to optimize for costs by getting access to cheaper credits)
The video editing is going to be a major skill requirement and those who embark on this could benefit largely. One can build a team of video editors that can be outsourced to create content. The team structure should be such that it leverages all the AI tools available out there.
PS:
- The AI slop that I created :) My-AI-Slop
- The GitHub repo Open Deep Research Video Creation