ai As a cinematographer, how should I confront AI technology, such as tools like ChatGPT? Midjourney?

[Chris Gu]

As we all know, ai technology is changing our daily life now when it was generally spread and used by people all around the world in the very beginning of 2023.

Some people even start to worry about whether their jobs will be replaced by ai in the future or not.

As for me, a young cinematographer, I've been thinking this question a lot of times.

It is interesting to listen to ChatGPT-4 and let's look how does it answer this question as below:

As a cinematographer, embracing AI technology like ChatGPT and Midjourney can greatly enhance your creative process and workflow. Here are some ways to confront and integrate AI tools into your profession:

Creative collaboration: Use AI tools to generate ideas, prompts, or even develop storyboards. These tools can provide you with fresh perspectives or unique concepts that you might not have considered before.

Pre-visualization: AI can help you visualize scenes or locations before shooting. With tools like Midjourney, you can create concept art or mood boards to better understand the desired look and feel of your project.

Time and resource management: AI tools can help you optimize your shooting schedule and resources, making your production more efficient.

Learning and development: AI technology is continually evolving, and staying updated with the latest advancements can help you understand and adapt to new techniques, tools, and trends in the industry.

Enhance communication: AI tools can help bridge the gap between different departments in a film production. For example, they can facilitate communication between cinematographers, directors, and other crew members by creating visual references or translating ideas into a visual format.

Post-production: AI tools can assist with tasks like color grading, visual effects, or even editing, saving you time and effort during the post-production process.

Adapt and evolve: Embrace AI technology as a tool to complement your skills and expertise rather than viewing it as a threat. AI can augment your creative process but cannot replace the artistic vision and intuition of a human cinematographer.

In summary, confronting AI technology in cinematography involves integrating these tools to enhance and streamline your creative process, stay updated with industry advancements, and optimize production efficiency. Embrace AI as a valuable resource that complements your skills and experiences, allowing you to create visually stunning and impactful projects.

What I write above is only for reference. I'm happy to talk about this with you guys!

[Uli Meyer]

It makes sense to embrace new technology but please consider this: everything you mention, i.e. storyboards, concept art, visual effects, colour grading etc are somebody else's job. AI will thin out the job market. For example,  it's a lot cheaper and faster to hire one AI wrangler for concept art than to hire a team of traditional artists. It's already happening and this is just the start. AI will be good for people at the top and bad news for many work-for-hire artists.

[Joerg Polzfusz]

Hi!

After being to a conference with AI as one of the topics, I played around with Dall-E 2 (attached: „fluffy robot in front of a spaceship“ according to Dall-E 2). So: How long will it take until an AI generates a complete 90min video on its own - 10 years or 20 years?

The fact is that there are already AI generated „photos“ used in commercials (e.g. by NorthFace, Nike…) and as „product images“ on some boxes (e.g. on the packaging of Kaufland‘s own brand). And there are „online shops“ (e.g. billiger.de) where an AI automatically mixes the manufacturer‘s descriptions with the user comments and ratings into the article’s description as shown on the webpage, reducing the marketing department’s work in this area by 80%.

However, so far, no AI is „creative“. E.g. all that Dall-E 2 does is „recycling“ existing images/photos, replacing single objects and maybe changing the style. So, if it would only be trained with images of candles, then it couldn’t invent an image of a light bulb. It also cannot invent a new drawing style on its own. In other words: When you are only doing dull/boring/repeating work, then you might be replaced by an AI within the next 20 years. But when you have to be really creative, then you’ll be save (of course only for some more years ?). However, you’ll be supported by AI soon. In fact, you already are - just think of the already existing auto-focus-cameras that detect humans in the image and that are automatically using these humans as reference points for the focus and then will follow these humans with the focus. Or think of all those digital still cameras that will automatically only take group images when everyone is smiling.

 

[Uli Meyer]

A friend of mine, Andrei Riabovitchev, (https://www.artstation.com/andrei) who is a brilliant traditional artist and has extensive credits as a concept artist and production illustrator, was one of the first to play around with AI imagery. The ones you can see here are word prompts only, no Photoshop, done in Mid Journey 5. When he started with AI, he took the results and used Photoshop and other programs to create compositions. By now the technology has advanced so far that all he has to do is feed it prompts to get the results you see here. And these only took minutes to produce.I believe that soon his name will becoe a prompt.

To think that AI is not going to thin out jobs in this industry is wishful thinking.

 

 

Edited March 24, 2023 by Uli Meyer

[Joerg Polzfusz]

Here’s an overview of some AI tools that can already turn text into video without any camera:

https://victorytale.com/best-ai-video-generators/

[Jon O'Brien]

What it might do is create a sense of competition. How can we make our picture warmer, more human, more interesting, more artistic, than the spirit in the machine?

[Chris Gu]

13 hours ago, Uli Meyer said:

A friend of mine, Andrei Riabovitchev, (https://www.artstation.com/andrei) who is a brilliant traditional artist and has extensive credits as a concept artist and production illustrator, was one of the first to play around with AI imagery. The ones you can see here are word prompts only, no Photoshop, done in Mid Journey 5. When he started with AI, he took the results and used Photoshop and other programs to create compositions. By now the technology has advanced so far that all he has to do is feed it prompts to get the results you see here. And these only took minutes to produce.I believe that soon his name will becoe a prompt.

To think that AI is not going to thin out jobs in this industry is wishful thinking.

 Sure things. I've been using Midjourney v5 for a few weeks, and there pictures below was made by it. Also, I only provided prompts without any post-production.

I would like to share some of my points with you all if you are interested in.

After experiencing MJ v5 for a while, let me summarize the biggest impressions.

There are two points.

Firstly, I believe those who have used it a few times would have the same feeling: at the current stage, it is difficult for users to control the scene very precisely through prompt words. It can be said that the result is in a vague state within the given prompts range. Just give AI more time.

Based on the understanding of the first point, the second point arises. How to obtain a scene that meets one's needs as much as possible? At present, it still lies in the prompts themselves.

From the view of the composition of the scene elements, the prompts should pay special attention to the overall logicality and accurate specificity. The latter is easy to understand, as the prompt words provide specific visible objects or scenarios. For example, you may say you want a romantic scene, but what is a 'romantic' scene? How does AI understand 'romantic'? Does it mean the same as what most people understand by 'romantic'? Instead of giving the prompt 'romantic' directly, I would recommend describing something like this: a French girl dancing in a forest under the backlight during the golden hour. This indeed requires imagination.

Then what does the logicality of the scene mean? I found that the more detailed the prompt words, the better they are not. I think AI's processing method is trying to give you what you need. However, the prompt words you give may be contradictory to AI. For example, you give "close-up shots, out-of-focus light spots" and a specific scene content "empty square". In my opinion, humans can understand these three descriptions, but putting them together for AI is contradictory because, logically, to shoot a close-up shot with out-of-focus light spots, environmental information is extremely lacking. However, you also provided AI with environmental information. In this process, it has to choose between the two, and the final result might be a close-up shot without knowing where it is, or a scene in the square, but not a close-up shot.

Please feel free to talk about this topic!

[Jon O'Brien]

I think it's regrettable that creative jobs are threatened by this new technology. Yes it looks like jobs in CGI and other forms of digital image creation and concept art are going to start disappearing or already have. But I find the new technology profoundly boring and wouldn't be interested in movies made using it. I like creative works made by real people and will pay to go and see those, if they are worth seeing. Perhaps that's just me but I think there is a rising movement out there that is rejecting all this digital crap being generated.

[Saikat Chattopadhyay]

On 3/25/2023 at 1:13 PM, Chris Gu said:

 Sure things. I've been using Midjourney v5 for a few weeks, and there pictures below was made by it. Also, I only provided prompts without any post-production.

I would like to share some of my points with you all if you are interested in.

After experiencing MJ v5 for a while, let me summarize the biggest impressions.

There are two points.

Firstly, I believe those who have used it a few times would have the same feeling: at the current stage, it is difficult for users to control the scene very precisely through prompt words. It can be said that the result is in a vague state within the given prompts range. Just give AI more time.

Based on the understanding of the first point, the second point arises. How to obtain a scene that meets one's needs as much as possible? At present, it still lies in the prompts themselves.

From the view of the composition of the scene elements, the prompts should pay special attention to the overall logicality and accurate specificity. The latter is easy to understand, as the prompt words provide specific visible objects or scenarios. For example, you may say you want a romantic scene, but what is a 'romantic' scene? How does AI understand 'romantic'? Does it mean the same as what most people understand by 'romantic'? Instead of giving the prompt 'romantic' directly, I would recommend describing something like this: a French girl dancing in a forest under the backlight during the golden hour. This indeed requires imagination.

Then what does the logicality of the scene mean? I found that the more detailed the prompt words, the better they are not. I think AI's processing method is trying to give you what you need. However, the prompt words you give may be contradictory to AI. For example, you give "close-up shots, out-of-focus light spots" and a specific scene content "empty square". In my opinion, humans can understand these three descriptions, but putting them together for AI is contradictory because, logically, to shoot a close-up shot with out-of-focus light spots, environmental information is extremely lacking. However, you also provided AI with environmental information. In this process, it has to choose between the two, and the final result might be a close-up shot without knowing where it is, or a scene in the square, but not a close-up shot.

Please feel free to talk about this topic!

Thanks for the lovely insight Chris,

also is it possible to share the prompts for 2nd or 3rd Image?

[Karim D. Ghantous]

All I can say about this is that I am not buying Adobe stock.

AI generated imagery is like free stock photos - high quality, and comprehensive enough.

I have been working on ways to authenticate camera files, particularly for the media and their customers. That to me is far more important. AI, or any art form, by definition, cannot replace a camera for documentary work. If you want to document the world you need a camera (or some kind of scanner). The key is to make it easy to authenticate those images or outputs. Shooting RAW files helps, but it's not the be-all and end-all, and it's not always the practical solution.

[Joerg Polzfusz]

Hi

Just two points:

* Indiana Jones 5 already contains AI generated scenes (basically all sequences that show a young Harrison Ford).

* You can use ChatGPT to automatically create transcripts of videos (and then e.g. to summarize it or in combination with tools like DeepL automatically translate it into other languages).

So we’ll all end up using AI in one form or the other pretty soon. Hence, I would recommend to play around with ChatGPT, Dall-e, Stable-Diffusion, FlexClip, Synthesia, … while it’s still free. ?

You’ll easily find out that the AIs have many limitations and will create a lot errors when generating images (humans with three arms and hence 18 fingers etc.). But the results are drastically improving every week. And you should at least know the current limitations when arguing against the use of AI in your next projects. ?

[Joerg Polzfusz]

https://share.synthesia.io/embeds/videos/db464d78-f10b-431c-b368-cca1aaeb4faa