Skip to main content

How to Prompt with Maestro?

Yug Saxena avatar
Written by Yug Saxena
Updated over 3 weeks ago

Introduction to Maestro

Maestro is a music generation model developed by Beatoven AI, designed to create instrumental tracks for background music in content creation. It leverages textual prompts to produce music tailored to specific genres, moods, and use cases, making it a powerful tool for content creators seeking royalty-free audio. This guide consolidates insights from observations, documentations, and advanced prompt engineering principles to help users craft effective prompts and maximize output quality.

How is it Different from the Last Beatoven.ai Model?

A key difference between an end-to-end generation model like Maestro and the previous stem-based approach lies in the creative process and the final output.

For Maestro, the user provides a text prompt, and the AI generates a single, complete, and polished instrumental track. This approach streamlines the workflow, making it incredibly fast and accessible for content creators who need ready-to-use background music without the need for production expertise. The creative input is focused on the prompt itself, and the model's "hallucinations" become part of the creative discovery, offering unexpected results that a user might not have conceived of on their own.

What works in Maestro?

Genres

Maestro reliably generates tracks in the following genres, based on user observations and documentation:

  • Latin

  • Jazz (especially '70s, '80s, '90s styles)

  • Blues

  • Reggae

  • Ambient

  • Retro (including Children's Music)

  • Wellness/Meditation/Focus/Sleep

These genres produce consistent outputs when paired with appropriate instrumentation and mood descriptors.

Instruments

  • Maestro is capable of single-instrument generations and multi-instrument tracks when instruments are described with clear, evocative adjectives (e.g., "bright, melodic piano," "gritty, distorted guitar").

  • Certain instruments are automatically generated well for specific genres, even if not explicitly requested (e.g., brass in Latin or Jazz).

  • Try using alternatives for terms that don’t seem to work. Here are some examples:

    • Replace "skanking guitar" with "rhythmic offbeat guitar" for Reggae.

    • Use "jazzy guitar" or "smooth piano" for accessible descriptors.

Functionalities

  • Single-Instrument Tracks: Maestro can generate tracks focusing on a single instrument, ideal for minimalist compositions.

  • Descriptive Prompting: Responds well to prompts that specify the role and character of instruments (e.g., "mellow piano leading the melody").

  • Parallel Generation: Opening two windows/tabs to generate four tracks simultaneously increases the likelihood of high-quality output.

  • Emotional Cues: Incorporating emotional stimuli (e.g., "evoke warm memories of the '80s") enhances creativity.

  • Arc Description: Describing a musical arc (e.g., "start sparse, build to a gentle climax") improves structural coherence.

What Doesn’t Work in Maestro?

Genres

  • Indian: Often misinterpreted as Oriental. (Note: A fine-tuned Indian music model is coming soon.)

  • Rock: Unreliable unless specified as subgenres like "grunge rock" or "classic rock."

  • Synth: Generic "synth" prompts yield inconsistent results; use "electronic" or specific synth types (e.g., "analog synth pads").

  • Genre Fusions: when combining two or more genres (i.e a core genre such as hip-hop, with another influencing genre, such as gospel) the generation gets lost and generates incoherent music.

  • Gospel: simple prompts of choirs and organs generate noise.

  • Country: When prompted to create different subgenres, it generates noise or reverts to the main style (polka rhythms with twangy guitars and banjos).

Instruments

  • Choice of words: Sometimes instrument names are needed to be re-phrased. For example, "rhythmic guitar" or “horns” may work better than like "skanking guitar" or "brass/trumpet" may not work as expected. Alternatives like "rhythmic guitar" or “horns” or relying on genre-specific defaults (e.g., brass automatically (un-prompted) in Jazz or Latin) are more effective.

  • Vocals: Maestro is not known to generate lyrical or non-lyrical vocals (e.g., humming, choirs), though this is expected in future updates.

  • Foley/Sound Effects: Coming soon, though not currently supported.

Functionalities

  • Duration: Tracks are capped at 2:30 minutes.

  • No Transitions: Maestro generally does not generate extremely dynamic transitions between musical sections.

  • No Editing: Generated tracks cannot currently be trimmed or edited post-generation.

  • No Stem Downloads: Individual instrument stems are not available, though this feature is expected soon.

  • Technical Terms: Specific musical terms like BPM, key, scale, or time signatures (e.g., "120 BPM," "C# minor," "6/8") lead to unreliable results.

Prompt Structure

An effective Maestro prompt may include the following components:

  1. Genre: Specify one broad musical category (e.g., Jazz, Reggae). Use subgenres for unreliable genres (e.g., "grunge rock" instead of "rock").

  2. Instrumentation: List all instruments with descriptive adjectives (e.g., "bright acoustic guitar strumming chords, mellow piano leading the melody"). Avoid technical or problematic terms.

  3. Mood: Use concise, evocative keywords (e.g., "calm," "nostalgic").

  4. Tempo: Use qualitative descriptors (e.g., "slow," "upbeat") instead of specific BPM values.

  5. Use Case: Define the context (e.g., "background for a travel vlog," "tense chase scene in a thriller").

  6. Duration: Note that duration is capped at 2:30 minutes; trim post-generation if needed.

  7. Additional Fields:

    • Video Context: Describe the video type (e.g., "montage of serene landscapes").

    • Film Scene: Specify a film scene type (e.g., "introspective moment in a drama").

    • Audio Tags: Up to 5 keywords for sonic textures (e.g., "reverb-drenched," "minimalist").

    • Energy: Set to "low," "medium," or "high" only if explicitly implied; otherwise, use null.

Best Practices

  • Start Simple and Iterate: Begin with a concise prompt and refine based on output (e.g., adjust "skanking guitar" to "rhythmic offbeat guitar").

  • Leverage Reliable Genres: Prioritize Latin, Jazz, Reggae, etc., for consistent results.

  • Align Mood and Use Case: Ensure mood complements the context (e.g., "tense, dramatic" for a thriller).

  • Avoid Technical Terms: Use descriptive language (e.g., "fast-paced" instead of "120 BPM").

  • Test Multiple Prompts: Generate four tracks in parallel to compare and select the best.

  • Apply Arc Description: Describe a musical arc (e.g., "start with a sparse piano melody, then add soft drums").

  • Use Emotional Cues: Add subtle emotional stimuli (e.g., "evoke warm memories") for creativity.

  • Validate Outputs: Evaluate against criteria like prompt adherence, instrumentation, mood, and quality and provide us feedback to help improve maestro.

Examples

Example 1: Jazz Study Playlist

Prompt: "Create a smooth jazz track with a soft piano leading a gentle melody and brushed drums, with a calm and nostalgic mood, suitable for a study playlist. Start sparse and build to a warm climax."

Example 2: Reggae Beach Vlog

Prompt: "Create an upbeat reggae track with light percussion, with a happy and relaxed mood, suitable for a beach vlog."

Example 3: Ambient Yoga Session

Prompt: "Create a meditative ambient track with soft pads, starting minimal and gradually layering instruments, with a calm mood for a yoga session."

Example 4: Evocative Piano Solo

Prompt: “Create a melancholic yet hopeful piano solo that evokes the feeling of watching the first light of dawn break through a misty window. The piece should begin with a sense of quiet solitude and gradually build in emotional intensity, introducing a fragile sense of optimism as the melody unfolds. The tempo should be slow and deliberate, with a focus on rich harmonies and expressive dynamics.”

Suggestion Box

To improve your Maestro experience, consider sharing feedback with Beatoven AI:

  • Request New Features: Advocate for adjustable track durations, vocal support, or dynamic transitions to enhance flexibility.

  • Suggest Genre Improvements: Highlight genres for better reliability in future updates.

  • Propose Instrument Additions: Recommend specific instruments for inclusion.

  • Share Prompt Ideas: Share successful prompts and tag us, to help refine Maestro’s algorithm and expand its creative potential.

  • Report Limitations: Note any issues with prompt adherence or output quality to inform future model enhancements.

Submit your suggestions via the Beatoven.ai Help Center or contact their support team to contribute to Maestro’s evolution.

Did this answer your question?