AI TOOLS COURSE

AI Tools

Alberto Sánchez Acedo | URJC y Prodigioso Volcán S.L.

Pasquale Cascarano | UNIBO

Nerea Benítez Aranda | URJC

Alejando Carbonell Alcocer | URJC

course Objectives:

• Approach to the AI concept
• Learning different types of AI
• Practical exercises with AI tools

1.1. AI CONCEPT

• AI began to be developed in the 1950s, but it was not until the 2010s that we had the necessary computing power. It is in full scale and, according to communications company 'Prodigioso Volcán', will reach its peak of acceptance and adoption over the next decade.

• 'Artificial Inteligence' term coined in 1956, at the University of Darmouth (USA).

1.1. AI CONCEPT

• Types:

• Weak AI: solve specific task without a general comprehension (Voice assistant)

• Strong AI: a smart character from a sci-fi movie (Matrix)

1.2. AI LEARNING

Machine Learning: 'machines or computers are able to learn autonomously through data analysis and pattern identification. Thanks to the knowledge they acquire, they then make predictions' (UOC, 2022).

Deep Learning: 'deeper level of machine learning as machines try to mimic the human brain' (UOC, 2022).

1.3. SOME TYPES OF GENERATIVE AI

Text to text: ChatGPT (https://chatgpt.com/)

Text to image: Adobe FireFly (https://www.adobe.com/products/firefly.html)

Text to video: Pika (https://pika.art/home)

Text to audio: Play.ht (https://play.ht/)

3D objects generation: Meshy (https://www.meshy.ai/)

Generated hybrid content: Vidnoz (www.vidnoz.com)

1.3.1 CHATGPT

• Text to text: ChatGPT is the artificial intelligence created by the company Open AI. This tool allows you to generate content from prompts.
• Web: https://chatgpt.com/

Examples prompt structure: [Context] + [Specific information] + [Purpose] + [Format of response] = Prompt perfect (BURJ, 2024) 'I request an image of size [specify 'large', 'wide', 'panoramic', square or 'portrait', etc], of [detailed description of the main subject], in an environment of [description of the environment or context], with an atmosphere that conveys [emotions or sensations]. The style of the image should be [artistic or technical style], and the predominant colours are [main colours]. Elements such as [specific additional elements] should be included' (BURJ, 2024).

Exercise

• Create your account in ChatGPT and ask ChatGPT to write a short description of a videogame

• Ask ChatGPT to create your main characters (physical and feelings) and introduce them in ChatGPT

• Ask ChatGPT to add some context information about your game

Results

Exercise

• Ask ChatGPT to generate a code to create a videogame

• Copy the code in Codepen.io (a web where you can visualize the code)

• Ask Chatgpt to generate a code modifying parameters to improve the videogame

• Share a screenshot of each result in the collaborative document and explain briefly in class the modifications you have made and if it resembles to your initial idea

Results

1.3.2 Adobe Firefly

• Text to image: this AI also offers the option to generate the images in different styles and aspect ratios (Sanchez-Acedo, 2023).
• Web: https://www.adobe.com/products/firefly.html

• Introduce your promt: 'A happy dog eating watermelon in the garden'

• Choose the aspect ratio, a reference image, the style or adjust the different parameters.

Exercise

• Use a simple prompt for a picture. Then use a complex promtp including more details and see if there is a difference. Use also different languages and compare the restults

• Generate an images of an animal or a landscape using the same prompt in Firefly (https://www.adobe.com/products/firefly.html) and Copilot (https://copilot.microsoft.com/)

• Create an avatar image (of you or a fictional person), as realistic as possible using both AI's

• Share your results and explain briefly in class your impressions with both AI's

Let´s animate our images

• D-ID tool

Results

1.3.3 PIKA

• Text to video: Pika allows users to generate video content from existing text, images or videos.
• Web: https://pika.art/

• Pika enables lip-syncing with written dialogue, application of sound effects, and canvas scaling.

Exercise

• Create a video sequence.

• Try and adjust the different parameters available such as sound effects or the style of the sequence

• Download the final result and upload it in the shared document

• Explain briefly in class your final result

Results

1.3.4 PLAY.HT

• Text to voice: voice Generator with 600+ AI voices that creates ultra realistic Text to Speech voiceovers (play.ht, 2024)
• Web: https://play.ht/

• Convert text to audio and download as MP3 & WAV files

• AI Voice Cloning

CO-PRODUCTIONNON-ATTRIBUTION ANONIMITY AUTHENTICITY

• Ultra-realistic AI voices

Exercise

• Create a new account in Play.ht2

• Generate a Virtual Voice in Studio

• Clone a famous voice and generate a new audio with this achieve

• Ask to ChatGPT to create a script for including it as a virtual voice.

Tips: you can try different voices, change spelling elements and adjust the parameters

Results

1.3.5 MESHY

• Meshy allows you to generate 3D objects from image or text.
• Web: https://www.meshy.ai/

• You can adjust different parameters

• Meshy exports in different 3D formats

Exercise

• Generate a 3D object from a prompt or an image and choose the style

• Upload to the shared document a screenshot of the generated model

• Download the 3D object in GLB format and save it to work with it in the A-Frame class.

Results

1.3.6 VIDNOZ

• This tool allows you to generate different types of content.
• Web: https://www.vidnoz.com/?test=page1

• Example Avatar Creation:

• Example AI Photo Tools:

CO-PRODUCTIONNON-ATTRIBUTION ANONIMITY AUTHENTICITY

• Example AI Voice Tools:

• Example AI Video Tools:

Exercise

• Choose at least one of each type of tool (imagen, video and audio) and create content of each one

• Download the final results and upload them with your name in the collaborative doucument

• Explain briefly in class the final results you have made

Results

AI Tools Course

Audiovisual and multimedia production:

This work was supported by the Autonomous Community of Madrid (Spain) with a grant for industrial doctorates (IND2022/SOC-23503) with the collaboration agreement with Prodigioso Volcán S.L; Universidad Rey Juan Carlos (ID 501100007511)