Creating a Voice Replica Using AI 0:09

• Josh Lomelino discusses the use of AI-powered voice synthesis to create a voice replica, emphasizing the challenge of matching human recordings.

• He highlights the effectiveness of using text prompts to quickly prototype, test, and revise scripts or generate finished audio files.

• Josh mentions his preference for the 11 labs tool, which offers a studio mode for producing longer form audio tracks.

• He shares his initial struggles with the tool and how contacting their support provided helpful suggestions.

Training the System for Consistent Output 1:24

• Josh explains the importance of training the system with a consistent audio sample to avoid unnatural variations in volume and tone.

• He describes his initial mistake of using diverse recordings from different sessions, which led to inconsistent results.

• Josh emphasizes the need for a controlled environment with a single, consistent audio sample for better results.

• He plans to demonstrate the settings that produce the best results for replicating his voice in the user interface.

Optimizing Generated Audio Files 2:56

• Josh advises generating audio sparingly to avoid exhausting monthly credits and recommends starting with smaller sections of text.

• He explains the process of refining the output and generating both wave and mp3 audio files for different applications.

• Josh mentions the importance of storing both wave and mp3 files for secure storage and project organization.

• He notes that it may take several attempts to develop a method that works well for the user.

Exporting and Integrating Audio Files 4:19

• Josh describes two methods for uploading audio files to Virtual avatars: exporting both wave and mp3 versions or integrating the 11 labs API directly with Hey Gen.

• He prefers using the wave audio file for higher quality and to avoid double compression but acknowledges the need to export the mp3 format for larger tracks.

• Josh explains the integration of the 11 labs API with Hey Gen, which allows for rapid development of prototypes and large volumes of content.

• He mentions the need to break up scripts into manageable sections for efficient processing by the software.

Automating Video Production with AI 6:02

• Josh discusses the ability to produce videos at scale by automating both audio and video avatars from text.

• He highlights the productivity gains from using AI to generate video scripts and produce audio and video automatically.

• Josh notes the cost of AI-generated voice and the strategy of using high-quality audio only when necessary.

• He explains the use of draft versions of scripts with Hey Gen's voice replica to refine the script without incurring additional costs.

Finalizing and Exporting Scripts 8:04

• Josh describes the process of finalizing scripts and either reading and recording them manually or using the 11 labs integration within Hey Gen.

• He mentions the use of a side-by-side display setup with a Google document and video avatar performance for quick edits.

• Josh emphasizes the usefulness of this method for high-end projects that require detailed polishing and iteration.

• He concludes the demo by encouraging the use of digital voice replicas to scale beyond time constraints and improve productivity.

Setting Up Lighting and Color Values 0:08

• Josh Lomelino explains the importance of setting up lighting and color values once to achieve consistent results over time.

• He emphasizes the need to test lighting and color values before batch producing a group of avatars.

• Josh mentions the flexibility to make further adjustments later using L, U, T color modifications or color correction tools.

• The workflow allows for the efficient production of 10 to 50 avatars, ensuring visual polish from the start.

Consistency in Project Settings 1:42

• Josh highlights the necessity of matching video editing project settings to the specifications of the recording camera.

• He provides an example of setting up a project for a Logitech 4k camera and ensuring consistency in frame size and frame rate.

• Josh advises checking file properties to extract frame size and frame rate if unsure.

• Consistency in project settings is crucial for mass producing different clips.

Creating a Master Sequence 2:59

• Josh sets up a master sequence to serve as a template for duplicating sequences as needed.

• He uses a clear numbering system for sequences, labeling each avatar with a specific outfit and camera angle.

• Examples include Avatar 001, DIRECT address, no hands, and Avatar 0013, quarter view.

• Josh organizes sequences in a dedicated folder called a bin for project organization.

Batch Rendering with Adobe Media Encoder 4:56

• Josh explains the process of adding clips to a Batch Render Queue using Adobe Media Encoder.

• He selects in and out points for each camera angle, creating dedicated files for each angle.

• Josh configures the encoder to render only the specified in and out range on the timeline.

• Each camera angle should be exported as an individual MP4 file, specifying the folder location and file name.

Finalizing and Organizing Project Files 6:40

• Josh emphasizes the importance of organizing project files, including original source files, rendered clips, and project files.

• He advises saving the video editing project frequently as a fail-safe for future edits.

• Josh highlights the need to review source footage for any performance anomalies and correct them.

• The workflow allows for the removal of outdated avatars and recreation without problematic movements.

Backing Up and Scaling Content Production 8:25

• Josh frequently backs up his entire project folder by compressing it into a zip file for disaster recovery.

• He mentions the time investment upfront to create polished assets and resolve hiccups.

• Josh advises starting with manual methods and gradually scaling to more advanced techniques.

• The well-organized project structure saves time, enables content production scaling, and supports high-performance results.

Creating an Automated Performance Using Audio 0:08

• Josh Lomelino explains two options for creating an automated performance: using a text-to-speech generated audio file or recording the performance using a high-quality microphone.

• He emphasizes that recording with a high-quality microphone yields the best results and will demonstrate this method in the demo.

• Josh mentions that the next demo will cover creating a fully automated performance using text, automating the entire process from audio capture to video production.

• He notes that while the automated process is efficient, it may not match the quality of a live performance.

Preparing and Exporting Audio Recordings 1:09

• Josh discusses the importance of using a high-quality audio file for the best results and mentions uploading the audio recording to a digital avatar.

• He explains the need to export an uncompressed WAV file and an MP3 file optimized for web use, highlighting the importance of having both options ready.

• Josh typically records his audio directly into Camtasia, which he finds to be the fastest way to capture high-quality audio for quick editing.

• He demonstrates how to export a local file and choose between saving it as a WAV or MP3 file, noting that other audio editing tools can also be used.

Generating Video Performance with Digital Avatar 2:29

• Josh explains the process of generating a video performance by dragging and dropping the audio file into the project and adjusting the start and end times of the digital avatar.

• He mentions exporting the production to render the performance into an MP4 file and downloading it into the project.

• Josh highlights the use of the chroma key or ultra key function to remove the background and seamlessly integrate the digital avatar into any backdrop.

• He provides examples of using this technique for reading from a script, repurposing audio from live presentations, and creating matching visuals with on-camera performances.

Combining Performance Modalities and Future Demos 3:54

• Josh discusses the challenges of managing all three performance modalities (screen recording, audio, and digital avatar) simultaneously and the importance of practicing beforehand.

• He explains how to export the audio from a demo, generate a digital avatar, and overlay it onto the video, showing the versatility of combining these elements.

• Josh mentions upcoming demos that will cover generating audio using generative AI from text alone, creating a fully automated workflow.

• He will also demonstrate automating the creation of slides and the precise timing of each slide's animation, allowing for a completely hands-free production system.

Creating a Fully Automated Performance from Text 0:08

• Josh Lomelino explains the process of creating a fully automated performance directly from text, including generating audio prompts using Otter AI.

• He describes how he brainstorms ideas while walking and exports the subtitle transcript file, SRT, to process it with AI tools like Claude or ChatGPT.

• Josh mentions breaking up long scripts into manageable blocks of 1800 characters and generating a year's worth of content for various platforms.

• He emphasizes the use of text, whether written manually or spoken and transcribed, to craft a video script using two primary methods.

Generating High-Quality Voice Clones 1:51

• Josh discusses creating a high-quality voice clone using 11 Labs, initially finding the results artificial but later perfecting the settings.

• He highlights the importance of using a consistent audio clip for training the voice digital double, ideally around three hours of spoken audio.

• Josh explains the challenges of recording consistently for three hours and how he stitches together previous demo recordings to create a large audio clip.

• He stresses the need for meticulous tracking of audio settings to ensure uniformity and avoid sudden changes in volume or tonal quality.

Optimizing Audio Recording for Consistency 3:36

• Josh shares his experience of recording multiple live sessions with an audience, which infused the audio with personality and energy.

• He explains the importance of having consistently dialed-in audio for generating a high-quality performance, as the AI listens to everything in the audio track.

• Josh mentions the time and cost involved in using 11 Labs, which can take up to six to eight hours to analyze a voice and build a model.

• He advises against using cheaper models, such as the multilingual version one model or turbo 2.5, and recommends upgrading to the multilingual version two model for better results.

Using Hey Gen for Cost-Effective Prototyping 5:35

• Josh introduces Hey Gen as an alternative for creating generative content when 11 Labs burns through credits too quickly.

• He explains how he trains Hey Gen on his voice by uploading a 10 to 15-minute audio clip and generates unlimited videos for free, depending on the subscription plan.

• Josh describes the process of creating prototypes, making real-time adjustments to the script, and rendering multiple takes.

• He mentions using his phone in split screen mode while walking to make adjustments on the fly and then copying and pasting the revised script into Hey Gen.

Switching Between Hey Gen and 11 Labs 7:44

• Josh explains how he can switch the voice in Hey Gen to the high-quality production voice in 11 Labs with a click of a button.

• He highlights the downside of using Hey Gen, which is the risk of losing all credits if there are issues with the audio track in the final video.

• Josh prefers using the Studio tool in 11 Labs for targeted editing, which allows regenerating just portions of the audio without redoing the entire clip.

• He mentions the benefit of being able to download the WAV file and MP3 file from the Studio tool in 11 Labs as a fail-safe.

Organizing Video Production Phases 9:21

• Josh describes his workflow of treating production as two phases: the cheap, free voice phase and the final phase.

• He explains the process of pasting the text directly into the Hey Gen editor, listening to the prototype, and resolving issues before creating a new file in Hey Gen.

• Josh organizes his videos into two folders: a prototype folder and a final folder, for easy organization of his methods.

• He mentions using the multilingual version two model for cost-effective throwaway tests and training his voice with Hey Gen for free prototyping.

Leveraging Digital Doubles for High-Quality Videos 10:34

• Josh shares how he uses his digital doubles to replicate a performance of his voice and generate a corresponding video composite.

• He explains how he creates a script using Otter AI during a walk, copies and pastes it into his automated workflow, and produces a high-end video with minimal effort.

• Josh highlights the benefits of this workflow, which allows him to deliver excellence without skipping a beat, even when small inconsistencies would have derailed the process before.

• He concludes by mentioning the next steps in the following videos, which will cover adding automated visual elements on screen behind the Virtual avatar.

Using AI Tools for Content Creation 0:09

• Josh Lomelino explains how AI tools help him capture ideas and generate content directly from brainstorming sessions.

• He uses Otter AI to record his thoughts verbatim, which he then exports as an SRT file for transcription.

• The SRT file contains every word spoken along with time codes, making it easy to generate a full video script.

• Josh leverages AI tools like 11 Labs and Hey Gen to produce audio and video content from the transcribed text.

Generating Video Scripts from Transcripts 2:00

• Josh describes the process of generating a video script from the transcribed text using AI tools.

• He explains the difference between having a clear plan and a vague notion for the script.

• The AI can capture random ideas and generate multiple scripts within the Otter AI application.

• Josh then uses tools like Claude AI or ChatGPT to expand and refine the generated scripts.

Collaborative Writing with AI 2:35

• Josh aims to create a video script that his digital double can read aloud, reducing the need for extensive editing.

• He explains the collaborative writing process between himself and AI tools to generate drafts and revisions.

• The ultimate goal is to use AI to create a polished video script without spending hours on manual editing.

• Josh emphasizes the importance of spending time to perfect the AI prompting process.

Workflow for Converting SRT Files 3:51

• Josh demonstrates the workflow for converting an SRT file into a video script using Otter AI and Notepad.

• He highlights the importance of checking the prompts document for time-saving methods.

• Josh explains two methods for creating video scripts: word-for-word transcription and general direction.

• He provides detailed prompts for ChatGPT to convert SRT files into 1800-character blocks.

Handling Rough Brainstorming Transcripts 7:40

• Josh discusses handling rough brainstorming transcripts that require more assistance from AI tools.

• He explains the need to be mindful of checking each word when using AI to generalize the transcript.

• Josh provides a prompt for ChatGPT to convert the SRT file into a video script and fix grammatical issues.

• He emphasizes the importance of ensuring the script is readable by the AI digital double.

Challenges with AI-Generated Scripts 10:06

• Josh mentions potential challenges with AI-generated scripts, such as mispronunciation by the digital double.

• He explains the time-consuming process of manually correcting AI-generated scripts.

• Josh introduces a prompt for a cleanup pass to automatically correct readability issues.

• He advises copying and pasting the corrected script into the video script document for backup.

Finalizing the Video Script 12:23

• Josh explains the final steps of rendering the script as a prototype using a free voice clone.

• He advises listening to the playback and adjusting the script for pronunciation issues.

• Once satisfied with the prototype, the final audio can be generated using tools like 11 Labs.

• The final audio clip can then be uploaded to a Virtual avatar software for the final on-screen performance.

Batch Processing Multiple SRT Files 13:21

• Josh highlights the option to bulk export multiple SRT files from the Otter AI app for time savings.

• He explains how this process can be applied to a whole folder of SRT files.

• This method allows for the creation of massive amounts of content quickly and easily.

• Josh concludes the demo by encouraging viewers to try the process for themselves.