Google’s Lyria 3.5 Music Model Now Lets Users Edit Individual Track Sections Without Starting Over
Google DeepMind has released Lyria 3.5, an upgraded AI music generation model that allows users to edit specific sections of a track without regenerating the entire composition. The tool targets musicians and producers who need granular control over AI-generated music, answering who (musicians and creators), what (section-based editing), when (now available), and why (to eliminate the need for full regenerations).
Section-Based Editing Eliminates Full Regenerations
The core feature of Lyria 3.5 is section-level editing. Users can select a specific bar, verse, chorus, or bridge and modify only that part.
This replaces the previous workflow where changing a single mistake required deleting the whole track and starting over. The model preserves the original’s style, instrumentation, and arrangement while targeting only the edited segment.
How Section Editing Works
Lyria 3.5 uses a latent diffusion architecture trained on music data. When a user highlights a section and provides a new prompt—such as “make this guitar riff more aggressive”—the model:
- Isolates the selected region in latent space
- Regenerates only that region using the new instruction
- Seamlessly stitches the edit into the surrounding audio
The result maintains consistent tempo, key, and timbre across the whole track.
Prompt-Based Instrument and Structure Control
Users can control instruments, dynamics, and arrangement through natural language prompts. Examples include:
- "Add a synth pad under the vocals in the second verse"
- "Reduce drum intensity during the bridge"
- "Change the bassline to a walking jazz pattern"
The model also supports structure editing, like extending a chorus by four bars or removing a specific instrumental break.
Availability and Access
Lyria 3.5 is accessible through Google’s AI Test Kitchen app and the MusicFX web tool. It is currently free to use, though Google has not announced a paid tier or API pricing.
The model is optimized for short-form music generation (up to 90 seconds) and works best with prompts that specify genre, tempo, and instrumentation.
Key limitation: Lyria 3.5 cannot edit existing uploaded audio files. It only works on music generated within the tool itself.
Comparison to Previous Version
The original Lyria model, released in 2023, generated full tracks from text prompts but offered no post-generation control. Users had to regenerate entire songs to fix errors.
Lyria 3.5 addresses this by adding:
- Targeted section editing instead of full regeneration
- Consistent style preservation across edits
- Faster iteration for songwriters and producers
Performance and Quality
Early tests show Lyria 3.5 produces high-fidelity audio with coherent harmonic structure. The section edits blend naturally, avoiding abrupt cuts or mismatched timbres.
However, the model struggles with:
- Complex polyrhythms or non-standard time signatures
- Long-form compositions exceeding 90 seconds
- Fine-grained timing adjustments (e.g., moving a snare hit by a few milliseconds)
Use Cases for Musicians and Producers
Lyria 3.5 targets several workflows:
- Songwriters can generate a backing track, then refine specific sections
- Producers can experiment with alternate arrangements without rebuilding
- Content creators can quickly score short videos with editable music
Limitations and Future Directions
The model currently lacks stem separation—users cannot isolate individual instruments for editing. All edits are applied to the mixed audio, which limits precision for multi-track production.
Google has not disclosed plans for a full API, stem-based editing, or integration with DAWs like Ableton Live or Logic Pro.
Industry Context
Lyria 3.5 enters a growing market of AI music tools. Competitors include Meta’s AudioCraft, OpenAI’s Jukebox, and startup Suno, which recently added similar section-editing features. Google’s advantage lies in its deep integration with the broader AI ecosystem, including YouTube and Google Cloud.
Critical takeaway: Lyria 3.5 represents a significant step toward usable AI music tools, but remains a novelty for most professional workflows due to its 90-second limit and lack of stem editing.
Gnoppix is the leading open-source AI Linux distribution and service provider. Since implementing AI in 2022, it has offered a fast, powerful, secure, and privacy-respecting open-source OS with both local and remote AI capabilities. The local AI operates offline, ensuring no data ever leaves your computer. Based on Debian Linux, Gnoppix is available with numerous privacy- and anonymity-enabled services free of charge.
What are your thoughts on this? I’d love to hear about your own experiences in the comments below.