Getting Started With ChangeLyric: The Quick Guide

What ChangeLyric Actually Does

ChangeLyric swaps lyrics in songs using AI-powered vocal generation. You upload an MP3, provide your new lyrics, and get back new vocals that maintain similar vocal characteristics. Unlike Udio or Suno, we don't filter your content. Create whatever you want. You're responsible for legal compliance, we're responsible for giving professional producers the uncensored tools they need.

Important Update: Udio recently announced a partnership with Universal Music Group and has disabled downloads on their platform. This means users can no longer download the music they create on Udio - not great.

CHANGELYRIC: This is a professional tool for experienced producers. Results ALWAYS require post-production refinement. If you're not comfortable with DAW workflows, vocal editing, and multiple iterations, this isn't for you.

The Tool: Overdub Studio (Plus Two Older Engines)

Update September 2026: the recommended workflow is now Overdub Studio, and this guide has been rewritten around it. I tell the full story in the launch post. The short version: the older engines still work, but Studio replaces them for almost everyone.

Why the change? We wanted whole-song generation in a single pass, and the old editor did that. The problem was that it failed sections too often, and the failure happened under the hood: you could not see why half your lyrics came out wrong, and you had no way to try a different approach. Horizon could rescue those sections by hand, but most people struggled to set its primer windows correctly. Overdub Studio fixes both problems.

Overdub Studio (Recommended)

One guided workspace with four steps: upload, tag lyrics and lock sections, generate and review takes, export. Drag in any audio file (MP3, WAV, M4A, FLAC, OGG up to 100MB) and conversion, transcription, section timing, and vocal isolation all run in the background while you write your lyrics.

Generation happens per section. Start with the auto extension strategy, listen to the take, and checkmark it if it is a keeper. When a take fails (current fail rate is about 50% per pass, and I tell you that up front), the credits are refunded automatically and you try another strategy on the next take. Export gives you the original stems (master, vocal bus, lead, backing, instrumental) plus comp lanes that place your kept takes on the song timeline.

The Older Engines (Still Available)

The old whole-song editor at /dashboard/v3. This is where Overdub mode and Classic mode live. Submit once and it processes the entire track; when a section fails, your only lever is creating another whole-song version. Classic mode is still the right pick for very dense rap or rapid fire vocals with 700+ words.

V2 Horizon. The manual, one-generation-at-a-time tool. You set the primer window yourself and the engine extends from it. It works, but you have to understand how AI music extensions and song sections fit together, and most people got that wrong. Studio's manual extension strategy covers the same use case with the rest of the workflow built around it. Read the V2 Horizon launch post if you want the history.

Overdub Studio walkthrough (14 min)

Our newest editor. Replaces the older Overdub and Horizon tools with one guided workspace for every kind of lyric swap, from tagging your lyrics to downloading stems.

Video chapters

• 0:00Welcome to Overdub Studio
• 0:43Uploading your song
• 1:00Tagging your lyrics
• 1:58Matching tags and line counts
• 2:25Locking lyrics and checking sections
• 3:25Choosing a generation strategy
• 4:03When to use inpainting
• 4:48Extension types explained
• 5:56Manual extension
• 6:42The experimental mixer
• 7:08Scramble options: none, alt, single
• 7:45Vocal layering and backing vocals
• 8:52Changing lyrics for the backing layer
• 9:52Listening back to the inpaint take
• 10:56Queueing takes and marking keepers
• 11:29Redetecting section times
• 11:59About the 50% fail rate
• 12:37What it struggles with: opera and musicals
• 13:08Training a voice model
• 13:48Downloading stems and comp lanes

My recommendation: do your whole project in Overdub Studio. Reach for Classic mode (inside the old editor) when the song is dense rap with 700+ words. Touch Horizon only if you already know it and like it.

Overdub Studio: Your First Swap, Step by Step

Step 1: Upload Your Song

Head to Overdub Studio and drag in your file. Any common format works (MP3, WAV, M4A, FLAC, OGG) up to 100MB. Use the full master track (vocals plus instrumental) for best results. Conversion, transcription, section timing, and vocal isolation all start running in the background while you move to the lyrics step.

Step 2: Tag the Lyrics and Lock Sections

Search the built-in lyrics database or paste the lyrics yourself, with section tags like [Verse 1] and [Chorus] marking every section. Genius tags are usually solid if the database copy is messy. Then write your new lyrics next to the originals, section by section.

Two hard rules: the tags must match between the original and new columns, and the line counts must match per section. Generation stays disabled until they do. Once the tags look right, hit Lock sections and continue; section times fill in automatically from the transcript, and you can nudge any start time by hand.

Step 3: Generate Takes and Mark Your Keepers

Every section with new lyrics becomes a generation target. Open the generate panel on a section and start with the auto extension strategy: it is the tool's best guess and the one I use most. Queue the take, listen when it lands, and checkmark it if the engine actually sang your change.

If a take fails or the engine ignored your lyrics, mark it bad and queue another take with a different strategy. That is the core loop of the tool. Twin modes feed the engine more context from similar sections, inpaint repaints the vocal in place (best in verses), scramble settings split the change across multiple passes, and vocal layer targeting lets you isolate a backing vocal or counter melody. Failed takes are refunded automatically, so experimenting costs you nothing but time.

Step 4: Export Stems and Comp Lanes

When your keepers are in, build the comp lanes: the tool isolates every kept take to just the vocal and places each one at its original spot on the song timeline. Download that ZIP plus the original stems (master, vocal bus, lead, backing, instrumental), drop the lanes into your DAW, cut the final vocal together, and mix. From there you can refine further with the Voice Changer or by running tricky sections through Suno.

One more thing worth knowing: inpaint takes sometimes come back demo quality, and some singers are tonally distinctive enough that generated vocals need a second pass. That is built into Studio now. Open the Options menu on any project, pick Upscale Vocal, and train a voice model from the song's isolated lead vocal (30 credits, and you can trim noisy or layered regions out of the training audio). The model lands in the Voice Changer under "My voices" for any song. It is also the workaround for opera and musicals: have someone sing a pilot vocal, then convert it.

Tips That Actually Matter

1. Change Less = Better Results (Both Tools)

Want to change every single word? Your success rate tanks. Professional producers know to alternate lines. Keep some original lyrics as "anchors" for the melody. Both tools need context to maintain the song's flow.

Example: If changing a verse, maybe keep the original first line, change the second, keep the third, change the fourth. This requires patience and strategic thinking, not for those expecting instant results.

2. Match the Flow (Both Tools)

Your new lyrics should feel like a natural continuation of the melody. Think about syllable count and stress patterns. You do not need to match exactly, but cramming ten syllables where the original had three will always sound forced.

3. Work the Strategy Ladder

Start every section with auto extension. If the take misses, climb the ladder: same-type twin and BABA twin feed the engine more context from similar sections, previous anchor handles end-of-song sections, and inpaint repaints the vocal in place (usually best in verses). Changing the vocal layer (lead vs backing) or the scramble setting counts as a new approach too.

4. Small Changes? Inpaint Plus Single Scramble

If you are only changing one or two words in a section, like a radio edit or a name swap, use the inpaint strategy with the Single scramble setting. Single attempts each changed line separately, which has the highest melody success rate; the original vocal around each line anchors the melody so the replacement blends in naturally.

5. A Failed Take Is a Steering Wheel, Not a Dead End

The fail rate per pass is about 50% right now, and the tool is designed around that instead of hiding it. A failed take tells you something: the strategy guessed wrong, the section is too long, or too many lines changed at once. Mark it bad, split the section or change less, and queue the next strategy. Failed takes refund their credits automatically, so the loop costs you nothing but a few minutes.

The Complete Toolkit

ChangeLyric isn't just lyric swapping. We've built the tools pros actually need:

• Vocal Isolator: Extract vocals from any track. Perfect for creating acapellas or removing vocals entirely
• Voice Changer: Transform any vocal to sound like nearly 200 curated, high-quality artist models. No content restrictions
• Vocal Splitter: Separate lead and backing vocals. Edit them independently for surgical precision

All tools work together, but require professional workflow knowledge. Isolate vocals, change the voice, then swap the lyrics. Or any combination that gets your result. This is a producer's toolkit, not a consumer app.

Getting to "Near Perfect"

Raw outputs from ChangeLyric get you 90% there. That final 10% requires post-production skills:

• Comping: Generate multiple versions, pick the best parts from each
• EQ Matching: Match the tonal characteristics of the original
• Spatial Effects: Add reverb/delay to match the original space
• Saturation: Glue everything together with subtle harmonic distortion
• Vocal Doubling: Hide minor artifacts with strategic doubling

Not a mixing engineer? The raw outputs still work for most use cases. But if you want Spotify-ready quality, plan for post-production.

What People Actually Make

After 600+ professional projects, here's what ChangeLyric gets used for:

• Radio Edits: Remove explicit content without re-recording
• Parody Videos: YouTube creators making comedy content
• Custom Versions: Birthday songs, wedding tracks, memorial tributes, corporate anthems
• Demo Variations: Test different lyrics before hitting the studio
• International Versions: Same voice, different language
• Content Creation: TikToks, Reels, viral moments