Generative AI and professional post-production: Future fantasy or imminent reality?

Generative AI and professional post-production: Future fantasy or imminent reality?

Brie Clayton March 8, 2023

1 Comment

ChatGPT is not the first generative AI application. It's not even the first generative AI application to be released by its creators at OpenAI, or the first natural language processing tool driven by AI technology. But it is acknowledged to be the most advanced AI chatbot to date, and it's release to the general public in November 2022 has escalated the conversation around the use of AI across all industries. Other generative tools like DALL-E, Midjourney and Stable Diffusion garnered their fair share of attention over the last year, but ChatGPT's ability to create pretty convincing written content, on almost any topic, based on natural language prompts, has captured global interest to an extent seldom seen before.

So, what are the potential implications of this new breed of AI application for the professional post-production industry? We enlisted resident head of technical architecture at Jigsaw24 Media, Graham McGuinness, and indisputable expert in generative AI, ChatGPT itself, to help us figure this out by answering five (seemingly) simple questions.

Question 1 : How is generative AI being used in professional post-production workflows today? What is being very effectively leveraged right now are AI capabilities like speech-to-text, automated tagging and facial recognition, says McGuinness, adding that object recognition or machine vision is also developing at a rapid rate and becoming phenomenally powerful. He attributes ever growing training datasets and ever improving object detection algorithms with driving these impressive results.

What makes generative AI a hot topic now, according to McGuinness, is that In 2023, we're at the unintended intersection of near unlimited computing power, massive data sets and open-source tooling' that allows these capabilities to become accessible to wider sectors of the population.

To this end, his team has already started experimenting with using ChatGPT to write code (as an intelligent assistant), debug webserver setups and help write comprehensive knowledgebase answers to support questions. According to ChatGPT, generative AI is currently being used to up-scale and restore low-resolution footage, for colour grading, to generate realistic visual effects and even to automate editing. In professional audio post-production ChatGPT lists noise reduction, automatic dialogue replacement, music composition and speech-to-text as generative AI capabilities.

However, when asked to provide specific examples of generative AI tools used for video editing and colour grading, ChatGPT's answer included products like DaVinci Resolve's neural Engine, FilmConvert Nitrate, Adobe's Autoframe and Avid's MediaCentral Editorial Management tool which prompted our next question

Question 2 : What is the difference between AI and generative AI? ChatGPT answered this question with Generative AI is a type of artificial intelligence that uses algorithms to generate new content, such as images, videos, and audio, that resemble existing examples. Unlike traditional AI, which is programmed to perform specific tasks based on predefined rules, generative AI has the ability to learn and create new content by itself.

This is an important distinction - because while most of the professional post-production tools that ChatGPT listed in the previous questions have powerful AI trained' capabilities, they don't create new content in the same way that an AI-based text-to-image program does, and so don't necessarily qualify as generative' AI applications. It's a point that we felt was worth clarifying with one more follow-up question.

Question 3 : Is it fair to say that most professional video and audio post-production tools don't use generative AI capabilities at the moment, but that certain products use machine learning algorithms to improve functionality? Yes, that is a fair statement, ChatGPT confirmed. While there are some professional video and audio post-production tools that incorporate generative AI capabilities, most professional tools focus on using machine learning algorithms to improve functionality and assist with various tasks in post-production. For example, tools such as DaVinci Resolve and Adobe's Creative Cloud utilize machine learning algorithms to analyse footage and audio recordings and make intelligent adjustments to improve the overall quality of the finished product.

This is not to say that impressive generative AI video tools aren't already available. Citing products like fylm.ai and ColourLab, McGuinness says Some of their image correction capabilities are staggering, but the real divider is what they can do in the professional realm. The results are not always consistent. Synthesized speech is another area that's made massive strides in recent times, but McGuinness believes that we're still a way away' from tools like these having a real impact on mid-to-high-end post-production workflows. This is, in part, because the process can be painfully slow. But that's not the only issue to consider

Question 4 : What are some of the potential issues with using generative AI tools in post-production? Even though ChatGPT was trained using a database of around 300 billion words, OpenAI has been up-front about the fact that it may occasionally generate incorrect information. And this is true of any generative AI application. As McGuinness comments, Any AI-driven process still needs to be assessed by humans to find errors. The net result is that tasks that deploy these tools will either be non-critical, or able to be corrected ahead of wider consumption. Many non-trivial tasks like translation and subtitling currently make use of

LINK:	https://creativecow.net/generative-ai-and-professional-post-production...
	See more stories from creativecow

Most recent headlines