Replay: The EASIEST way to create AI Cover Songs!

Bob Doyle Media
26 Mar 202419:03

TLDRThe video script introduces Replay, a free AI-powered tool for voice conversion in music tracks. The host demonstrates how to use Replay to change the vocals of any song to a desired voice, showcasing its ease of use and the variety of voice models available. The process involves downloading the tool, selecting a song, choosing a voice model, and converting the track. The host also discusses advanced features like adjusting pitch, remixing with different models, and creating music from text prompts. The tool's ability to separate vocals and instruments is highlighted, as well as its potential for creative experimentation in music production.

Takeaways

  • 🎶 The channel focuses on exploring creative uses of AI, particularly in voice cloning and music conversion.
  • 🔄 'Replay' is a tool introduced for voice conversion, allowing users to change any song's voice to a desired one.
  • 🆓 Replay is entirely free to use, with no subscriptions required, making it accessible for everyone interested.
  • 📚 The tool works by downloading models from the internet, which are used for voice conversion, and most processing happens locally on the user's machine.
  • 🚀 Users can create their own songs or convert existing ones, including downloading tracks from YouTube, with Replay.
  • 🔍 The 'Weights' website is where users can find and download various voice models to use with Replay.
  • 🎵 The process involves selecting a song, choosing a voice model, and then converting the song to that voice.
  • ⚙️ Settings allow users to adjust the relative pitch and instrument pitch to better match the original song's vocal range.
  • 💻 The speed of conversion depends on the user's hardware, particularly the GPU used.
  • 🔗 Users can download individual tracks for remixing purposes, offering flexibility in audio editing.
  • 🎉 A unique feature of Replay is the ability to merge multiple voice models to create a new, unique voice.
  • 📝 The tool can also convert speech to different voices, not just music, providing a wide range of creative possibilities.

Q & A

  • What is Replay and how does it relate to voice conversion?

    -Replay is a tool for voice conversion that allows users to take any song and change its voice to any other voice they choose. It simplifies the process of voice cloning and music conversion, making it accessible for users interested in creative audio manipulation.

  • Is Replay free to use?

    -Yes, Replay is entirely free to use. There are no subscriptions or costs associated with the project, making it an attractive option for those looking to experiment with voice conversion without financial commitment.

  • What platforms is Replay available for?

    -Replay is available for download on Windows, Mac, and Linux platforms, providing a wide range of users with access to its voice conversion capabilities.

  • How does Replay handle the process of downloading models for voice conversion?

    -Replay occasionally downloads models from the internet while running, which may slow down the process. However, these model downloads are a one-time operation, and users can access a library of over 20,000 models on the Weights and Biases platform to enhance their Replay experience.

  • Can Replay be used to create music from a text prompt?

    -While Replay does have a feature that allows creating music from a text prompt, it is not as efficient or high-quality as other tools like Sunno. The text-to-music feature is more suitable for creating short audio snippets rather than full songs.

  • How does Replay separate the audio track for voice conversion?

    -Replay separates the audio track by first extracting the vocal track from the original song. It then allows users to apply different voice models to the vocals, creating a new version of the song with the desired voice.

  • What is the process of downloading and using a voice model in Replay?

    -To use a voice model in Replay, users download a ZIP file containing the voice model files from the Weights and Biases platform. They then extract these files into a folder and rename them for clarity. The model files are added to Replay by dragging and dropping the .pth file into the application.

  • How can users adjust the pitch of the voice and music track in Replay?

    -Users can adjust the pitch of the voice by changing the 'Relative Pitch' setting, which can be increased or decreased to match the desired vocal range. The 'Instrument Pitch' setting allows users to transpose the music track to better align with the vocal track's pitch.

  • What is the 'Multimodel' feature in Replay?

    -The 'Multimodel' feature in Replay allows users to select multiple voice models to use for a single song. This enables batch processing of voice conversion, creating a song with different voice models applied in sequence or simultaneously.

  • How can users merge models in Replay to create a new voice?

    -After selecting the 'Multimodel' option, users can choose two different voice models and then click on 'Merge' to create a new voice model that is a blend of the two selected models. This can result in a unique vocal sound that combines characteristics of both models.

  • Can Replay be used for converting speech as well as music?

    -Yes, Replay is not limited to music conversion. It can also be used to convert speech, allowing users to apply voice models to spoken words, creating a different voice for the speech without altering the background audio.

  • What are some of the advanced features of Replay that can enhance the user's creative process?

    -Replay offers advanced features such as the ability to download individual tracks (vocals, instrumentals) for remixing, the option to adjust the relative pitch and instrument pitch for better voice and music alignment, and the ability to create duets or unique vocal combinations using merged models.

Outlines

00:00

🎤 Introduction to Voice Conversion with Replay

The video begins with a channel introduction focusing on creative AI applications, particularly voice cloning and music. The host expresses excitement about a tool called Replay, which facilitates voice conversion in songs. It's highlighted that Replay is free and works offline, with occasional model downloads. The process involves downloading a track, using Replay to convert vocals, and selecting from a vast library of voice models. The host demonstrates the tool with a generated song and guides viewers on how to find and use models from Weights and Biases (W&B).

05:01

🔄 How to Use Replay for Voice Conversion

The host provides a step-by-step guide on using Replay, including downloading and renaming voice models, adjusting settings for voice conversion, and changing the pitch to match the original song's vocals. The video demonstrates how to convert a song's vocals using different models, such as Dean Martin and Squidward from SpongeBob. It also shows how to download source tracks for remixing and how to clean up noise in vocal tracks using audio editing software.

10:02

🎼 Advanced Features of Replay: Multimodel and Text-to-Music

The video continues to explore advanced features of Replay, such as the multimodel function, which allows the conversion of a song using multiple voice models simultaneously. The host also experiments with merging models to create new voice combinations. Additionally, a feature for creating music from a text prompt is discussed, though it is noted as being somewhat obsolete compared to other tools like Sunno. The host compares the output of Replay's text-to-music feature with Sunno's custom mode, favoring the latter for quality and efficiency.

15:03

📈 Final Thoughts and Call to Action

In the conclusion, the host reflects on the fun and creativity enabled by Replay, encouraging viewers to experiment with the tool and share their creations. The video ends with a humorous call to action for viewers to subscribe to the channel, with a playful threat of pursuit if they do not.

Mindmap

AI and Music Exploration
Voice Cloning and Music Interest
Replay as a Tool for Voice Conversion
Introduction to Replay
Free and No Subscription Required
Cross-Platform Availability (Windows, Mac, Linux)
Local Machine Processing
Features and Benefits
Downloading and Installing Replay
Downloading Models for Voice Conversion
Uploading or Recording Audio Tracks
Selecting and Applying Voice Models
Adjusting Settings for Voice and Instrument Pitch
Creating and Downloading Converted Songs
Process of Using Replay
Generating Original Songs for Conversion
Avoiding Copyright Issues
Integration with Sunno
Accessing Over 20,000 Models on Weights G
Searching and Downloading Specific Voice Models
Previewing Models with Speaking Samples
Exploring Voice Models
Remixing Songs with Different Voices
Multimodel Feature for Combining Voices
Adjusting Voice Ratios in Merged Models
Creating Duets and Unique Voice Combinations
Creative Applications
Using Audio Editing Software for Further Tweaks
Adding Effects and Adjusting Track Levels
Silencing Noise in Vocal Tracks
Post-Conversion Editing
Creating Music from Text Prompts
Limitations on Song Duration and Quality
Additional Features
Personal Enjoyment and Time Spent Using Replay
Invitation to Subscribe for Similar Content
Encouragement for Creative Exploration
User Experience and Recommendations
AI Cover Song Creation with Replay
Alert

Keywords

💡Replay

Replay is a software tool for voice conversion that allows users to change the voice in any song to a different voice of their choice. It is central to the video's theme as the host demonstrates how to use Replay to create AI cover songs. The host mentions Replay's ability to download models from the internet for various voices, which is a key feature in the process of voice conversion.

💡Voice Cloning

Voice cloning refers to the process of replicating a person's voice using AI technology. In the context of the video, voice cloning is an ongoing interest of the channel, and Replay is used to achieve a form of voice cloning by changing the original vocals of a song to a different voice.

💡Sunno

Sunno is a platform for generating music, which the host uses to create an original song for the purpose of the demonstration. It is relevant to the video as it provides a source for the original music track that will have its vocals replaced using Replay.

💡Models

In the context of Replay, models are voice profiles that users can download and apply to convert the vocals of a song. The host discusses downloading models from Weights & Biases, a platform that hosts these voice profiles, to use with Replay for voice conversion.

💡Weights & Biases

Weights & Biases is a platform where users can find and download voice models for use in Replay. It is mentioned in the script as the place where the host finds and downloads voice models like Dean Martin's to use in the voice conversion process.

💡Vocal Separation

Vocal separation is the process of extracting the vocal track from the instrumental part of a song. Replay automates this process, making it easier for users to replace the original vocals with a different voice, which is a significant part of creating AI cover songs as demonstrated in the video.

💡Relative Pitch

Relative pitch refers to the perceived height of a sound or voice and is an adjustable setting in Replay. The host uses it to match the pitch of the converted voice to the original music track, ensuring that the new voice sounds harmonious with the instrumentals. For instance, when converting to a higher pitched voice like Billie Eilish's, the host increases the relative pitch.

💡Multimodel

Multimodel is a feature in Replay that allows users to select multiple voice models to be used for a single song. The host demonstrates this by creating a song with both Darth Vader and Garth Brooks' voices, showcasing the versatility of Replay in voice conversion.

💡Remix

In the context of the video, remix refers to the process of re-recording a song with a different voice using Replay. The host uses the term when showing how to apply a new voice model to the original instrumental track of a song created with Sunno.

💡Text-to-Music

Text-to-music is a feature within Replay that generates short snippets of music based on a text prompt. Although the host mentions it is somewhat obsolete with tools like Sunno, it is still an interesting capability of Replay, allowing users to create music from a textual description of a style or genre.

💡AI Cover Songs

AI cover songs are songs that have been recreated using artificial intelligence to change the original vocals to a different voice or style. The entire video is centered around this concept, as the host explores the ease with which Replay can be used to make AI cover songs, transforming any song's voice to another of the user's choosing.

Highlights

Replay is a tool for voice conversion, allowing users to change any song's voice to a voice of their choice.

Replay is available for free, with no subscriptions required, and can be downloaded for Windows, Mac, and Linux.

The tool can download models from the internet, which are used for voice conversion and are typically a one-time operation.

Users can generate their own songs or convert existing songs, including those from YouTube, with Replay.

Over 20,000 voice models are available on Weights G, including singers and characters from popular media.

Replay allows users to audition different voice models directly within the application.

The voice conversion process can be adjusted for relative pitch to match the original song's vocal range.

Users can change the pitch of the instrumental track to harmonize with the converted vocal track.

Replay separates the audio track and allows for easy auditioning of multiple voices.

The converted tracks, including vocals and instrumentals, can be downloaded for further editing.

Replay enables remixing songs with different vocal models, creating unique versions of songs.

The multimodel feature allows users to apply multiple voice models to a single song for batch processing.

Merging models creates an entirely new voice model, blending characteristics of the selected models.

Replay can also convert speech to different voices, not just music.

The program includes a feature to create short snippets of music from a text prompt, although it's not as high quality as other methods.

Replay offers a high-quality separation of vocal tracks, which sound impressive when isolated.

The tool is highly engaging, allowing users to spend hours converting and remixing songs with different voices.

Replay enhances the capabilities of song creation platforms like Sunno, enabling users to use any singer's voice in their compositions.