What is MiniMax Music 1.5?
MiniMax Music 1.5 is a specialized generative music model provided by MiniMax. It is intended to create complete songs with vocals, lyrics, instrumental arrangements, and recognizable musical sections. MiniMax announced the model on September 11, 2025, presenting it as a next-generation music-generation system capable of producing tracks up to four minutes long.
The model is best understood as an audio-generation system rather than a general-purpose AI assistant. Its primary output is generated music, including vocal performances and accompanying instrumentation. It is not described as a text, image, video, speech-transcription, embedding, or general reasoning model.
Within the MiniMax catalog, Music 1.5 belongs to the provider's specialized MiniMax Music family. It should not be confused with later MiniMax music models or with MiniMax Music Cover, which is a separate offering for cover-style transformations. Those products may serve related creative workflows, but the supplied documentation does not establish that they are interchangeable with Music 1.5.
How the model generates songs
Natural-language music prompts
Music 1.5 accepts natural-language descriptions of the intended musical result. A prompt can describe characteristics such as musical style, mood, scene, or the type of project for which the track is needed. This makes the model suitable for users who want to begin with a creative brief instead of composing every musical element manually.
For example, a creator could describe the desired atmosphere for a short-video soundtrack or request a vocal song with a particular emotional direction. The available research does not specify a fixed prompt syntax or a required list of musical parameters, so users should not assume that every production control is exposed as a formal API field.
Structured lyrics and song sections
An advanced workflow can provide lyrics arranged around named song sections such as Intro, Verse, and Chorus. This gives the user more influence over lyrical placement and the overall progression of the composition than a purely open-ended prompt.
MiniMax emphasizes clearer differentiation among major sections, including intros, verses, choruses, bridges, and outros. In practical terms, this means the model is intended to produce a recognizable song shape instead of a track that remains musically uniform from beginning to end. The documentation describes this as a model focus and provider capability claim, not as a published benchmark result.
Main capabilities
Full-song generation up to four minutes
The clearest stated output limit is a maximum song length of up to four minutes. That duration makes Music 1.5 relevant to complete demos, background tracks, short-form video music, and other projects that need more than a brief musical sketch.
The four-minute figure is a maximum stated song length, not a guarantee that every generation will produce a complete, commercially finished master. The supplied research does not provide a separate specification for output file formats, audio resolution, download limits, or the number of generations allowed in a request.
Vocals and layered instrumentation
MiniMax describes Music 1.5 as capable of producing natural-sounding vocals with varied timbres and singing styles. It also highlights layered instrumental arrangements and dynamic playing techniques. The model supports both vocal and instrumental creative directions, although the research does not define a separate control panel or parameter set for selecting every voice or instrument.
Another stated capability is support for music featuring traditional Chinese instruments. This may make the model especially relevant for projects seeking instrumentation or musical color beyond common Western pop arrangements, although the supplied sources do not list the specific instruments supported.
Coherent musical structure
Music 1.5 is designed to distinguish major song sections and make transitions among them more coherent. This matters when the goal is a song with a verse-and-chorus progression, a bridge, or a deliberate opening and ending. It is a different objective from generating a short loop or a single ambient texture.
Inputs, outputs, and technical limits
| Specification | Available information |
|---|---|
| Primary input | Natural-language music descriptions; optional section-based lyrics |
| Primary output | Generated music audio |
| Vocals | Supported according to MiniMax's description |
| Instrumental music | Supported |
| Maximum stated song length | Up to four minutes |
| Text output | Not a stated output modality |
| Image or video output | Not supported or described for this model |
| Context window | Not published for this exact model |
| Maximum output tokens | Not applicable or not published for this audio-generation model |
| Structured output | Not documented as a general structured-data feature |
The model's documented input and output profile is narrow: text-based creative instructions and optional lyrics are used to produce music audio. The available research does not verify audio-input, image-input, video-input, tool-use, function-calling, streaming, caching, batch-processing, or fine-tuning support.
There is also no published knowledge-cutoff date, context-window size, or token limit for Music 1.5. Those omissions are important because language-model specifications do not automatically transfer to a specialized music model.
API availability and pricing
MiniMax announced a global API for Music 1.5 when the model launched. However, the current supplied documentation does not verify that the exact Music 1.5 endpoint remains available through a standalone pay-as-you-go API. MiniMax's Token Plan documentation states that music models were scheduled to stop being available through the Token Plan on August 20, 2026.
Accordingly, Music 1.5 should be treated as a legacy or availability-uncertain model rather than one with confirmed current API access. The Token Plan status does not by itself prove that every possible access route has been shut down, but no alternative current route is verified in the supplied research.
No verified price for input, output, generation, subscription access, or API usage is available. Developers should confirm the model identifier, endpoint, account eligibility, and current billing terms directly with MiniMax before designing a production workflow around it.
Strengths and trade-offs
Music 1.5's main strength is its focus on complete song creation. A single model is intended to handle lyrics-to-song workflows, vocals, instrumental layers, and recognizable song structure. That is more directly aligned with songwriting and soundtrack production than a general audio tool intended only for effects or short clips.
Its four-minute maximum is also useful for projects that need a relatively substantial track. Structured lyrics can provide more control than an entirely free-form prompt, while the stated emphasis on section transitions may help users create compositions with a conventional musical narrative.
The principal trade-off is uncertainty rather than a documented quality weakness. There are no supplied benchmark scores, published latency figures, confirmed prices, or verified current standalone API details. The model also does not provide the broad modalities or tool capabilities expected from a general-purpose AI system. Users seeking text reasoning, coding, image generation, video generation, transcription, or voice-only synthesis should choose a different type of model.
Best use cases
- Generating complete vocal-song demos from a creative brief and lyrics.
- Creating instrumental or vocal background music for short videos, films, games, and social content.
- Exploring arrangements before hiring musicians or producing a final recording.
- Developing virtual-idol, entertainment, or branded music concepts.
- Testing different moods, song structures, vocal styles, and instrumentation quickly.
- Creating music that incorporates traditional Chinese instrumental elements.
These use cases reflect the model's stated design and are most practical when the user can review, edit, or regenerate the resulting audio. The available research does not establish commercial licensing terms, so rights and usage permissions should be checked separately before publishing generated tracks.
When to choose MiniMax Music 1.5
Choose Music 1.5 when the central requirement is a complete vocal or instrumental song of up to four minutes, especially when natural-language direction and section-based lyrics are more important than detailed manual composition controls. It is a reasonable fit for ideation, demos, soundtrack drafts, and entertainment content when access is available.
Another music-generation option may be more appropriate if you need a currently confirmed endpoint, transparent usage pricing, audio-input transformation, a documented streaming interface, fine-tuning, or precise production controls. A later MiniMax music model may also be worth investigating for a new project, but the supplied research does not provide enough information to claim that a specific successor is better or currently accessible.
For non-musical work, Music 1.5 is the wrong category of tool. Use a language model for reasoning or code, an image model for visual content, a video model for moving images, and a dedicated speech system for transcription or voice synthesis. Music 1.5's value comes from its narrow specialization in generating song-like audio, not from general-purpose AI capabilities.
Bottom line
MiniMax Music 1.5 is a specialized song-generation model built around four-minute vocal and instrumental compositions. Its distinguishing features are natural-language music prompting, optional structured lyrics, layered arrangements, varied vocals, traditional Chinese instrument support, and an intended ability to maintain clearer song sections. Those capabilities make it relevant to creative audio production, but current access and pricing are not verified, and MiniMax documentation places music models outside the Token Plan from August 20, 2026. Treat it as a historically documented or availability-uncertain option until MiniMax confirms a usable current endpoint and commercial terms.