ChatVideoMusic

Model developer guide

MiniMax

15 website model guides

MiniMax spans text reasoning, Hailuo video and music, but those are separate integrations. Choose the deliverable first, then inspect the exact model route. In particular, the audited music interface shares its prompt and lyrics text, so a normal descriptive prompt may be sung instead of treated only as direction.

ChatVideoMusic

Practical guidance

Where should I start?

Editorial starting points, not a performance ranking. Check current app access before choosing a model.

Start with original song words

WaveSpeed documents vocals, accompaniment and section-tagged lyrics for Music 2.6. For this app, begin with concise original lyrics because the owner audit says the main input is also sent as lyrics.

Tradeoff: The app does not expose the gateway's independent style and lyric fields. A phrase such as 'soft guitar' may become sung text when mixed into the main input.

MiniMax Music 2.6 guide

Compare the Music 2.5 route

Replicate documents Music 2.5 as a vocal-song model with separate lyrics and style-prompt fields. Its distinct route is also the owner-audited fallback behind the ElevenLabs Music choice.

Tradeoff: The app sends the same text to both fields. Do not expect the gateway's independently controlled lyric/style workflow or infer how often fallback occurs.

MiniMax Music 2.5 guide

Technical details

What does the provider document?

Official documentation describes the named model or gateway—not every feature available in this app.

Do not inherit a newer model's thinking settings

MiniMax's compatibility reference distinguishes M3 from M3.1-Flash-Preview: tunable low-to-max reasoning_effort is effective for the latter. M3 supports adaptive or disabled thinking, while the documented M2.x models cannot disable it. The exact model ID and endpoint format matter when interpreting a default.

Two song gateways, different schemas

WaveSpeed Music 2.6 documents prompt, lyrics, is_instrumental, bitrate and sample_rate, but no numeric duration field. Its prose and schema disagree about whether lyrics are required. Replicate Music 2.5 separates lyrics of 1–3,500 characters from a style prompt up to 2,000. Neither schema makes the app's shared 500-character input equivalent to independent controls.

Complete website coverage

Explore all 15 model guides

Every existing catalog entry for this provider is linked below. Status labels come from the individual website guides; catalog coverage is not a guarantee of current app access. Historical and snapshot pages remain references, not extra active models.

Practical guidance

From a brief to a useful result

Original prompt examples, not generated-output demonstrations. They have not been generation-tested.

Workflow 1

Review a confusing interface failure

Provide a short bug description, redacted logs and a screenshot if supported. State expected behavior and keep the first review limited to one interaction.

Compare the reported checkout failure with the supplied screenshot and error log. List what is directly observed, what might explain it, and what remains unknown. Cite a visible element or log line for each claim. Propose the smallest three tests that distinguish a validation issue from a server error. Do not invent requests or claim to have run the application.

Check the result: Make sure the diagnosis respects the evidence and that each suggested test has a concrete expected outcome.

Use the M3 guide

Workflow 2

Write a compact original chorus

Use original words you would be comfortable hearing sung. Keep the full input under 500 characters, because the app has no separate custom-lyrics input; section tags are a structure request, not a timing guarantee.

[Verse] We fold the maps and leave the rain A little hope is ours again The road is quiet, the sky is wide We take the morning side by side [Chorus] Carry the light, carry it home No one has to walk alone One open door, one steady flame Tomorrow knows us by our names

Check the result: Listen for the intended words, intelligibility and section changes. Measure the resulting duration; do not assume the app's internal minute request controls song length.

Read the Music 2.6 guide

Technical details

What changes inside EZ Ai Assist?

Official provider and gateway documentation was reviewed for the contracts described below. App music behavior comes from the owner-supplied October 9, 2026 audit, not independently inspected application code or generation tests. A documented endpoint does not prove the app successfully uses it.

Owner music audit · October 9

The main prompt becomes lyrics

For Music 2.6, 2.5 and 1.5, the main prompt is sent as both prompt and lyrics. There is no separate custom-lyrics input or instrumental toggle. The app accepts 1–500 characters, which can conflict with the narrower Music 1.5 prompt range described in its guide.

Owner music audit · October 9

The requested minute is not forwarded

The client requests durationSeconds: 60 with no user-selectable duration. The audited MiniMax routes do not forward that duration or quality overrides. Music 2.6 uses WaveSpeed; 2.5 and 1.5 use Replicate. No 60-second output or version-wide maximum is guaranteed here.

Existing website guides · not live-tested

Chat and Hailuo need separate checks

Chat settings were not inspected for this update. Existing Hailuo guides report AIML routing; Hailuo 2.3 Fast is an audited flag on the same model ID, not independent proof of a faster provider endpoint. Music routing and control evidence must not be applied to video or chat.

Before you rely on a result

Limits, costs and tradeoffs

  • Lyrics supplied as descriptive instructions may be performed as words. This is a documented app-boundary warning, not a promise that the provider will reinterpret them correctly.
  • Provider instrument-only and quality fields are not necessarily app controls. The music gateway schema conflict and actual acceptance of short inputs remain unresolved without an authorized generation test.
  • Model speed claims, picker estimates and provider rates are not customer charges or measured app performance. This guide does not change billing, routing or fallback behavior.

Provider references and picker estimates are not EZ Ai Assist customer charges. Read the website pricing explanation.

Compare by task

Other paths to explore

Suno

Consider its audited dedicated custom-lyrics interface when independent lyric entry is central to the task.

Google Lyria

Compare prompt-led music with clear version-specific vocal and duration distinctions.

AIML

Understand why the ElevenLabs Music selection may fall back to MiniMax Music 2.5.

Common questions

MiniMax, without the assumptions

Why might MiniMax sing my instructions?

The owner audit says the app copies the same main prompt into both prompt and lyrics. A style instruction can therefore enter the lyric field. There is no separate custom-lyrics input in this integration.

Can I choose exactly one minute?

No. The music client requests 60 seconds internally, but the audited MiniMax handlers do not forward it. Neither a prompt requesting one minute nor a model card establishes exact returned length.

Is M3 the same model as Music 2.6 or Hailuo?

No. M3 is a text-output model with visual understanding. Music and Hailuo use separate model families, endpoints, gateway contracts and app controls. Their shared developer name does not combine the capabilities.

New to the terminology?

Official documentation

Sources & review notes

Highlighted provider contracts reviewed . Individual model guides carry their own review dates. Sources can change; these pages do not perform live health checks.

Official provider and gateway documentation was reviewed for the contracts described below. App music behavior comes from the owner-supplied October 9, 2026 audit, not independently inspected application code or generation tests. A documented endpoint does not prove the app successfully uses it.