MakeARap

The AI behind MakeARap

MakeARap brings together several AI models, each chosen for one job. Here is exactly which models we use, what each one does and how your words are handled.

Lyrics: you choose the writer

Pick a lyric writer in the studio with the model icons next to the language menu.

  • Claude Sonnet 5.5

    Developer: Anthropic

    Anthropic’s Claude Sonnet 5.5 is our default writer. If it is busy, Claude Opus 5.5 can step in.

  • DeepSeek V4.1 Flash

    Developer: DeepSeek

    DeepSeek V4.1 Flash, the fast model from DeepSeek. Free accounts write with it.

  • Grok 4.7

    Developer: xAI

    Grok 4.7 from xAI.

  • Gemini 4

    Developer: Google

    Gemini 4 from Google.

Free accounts use DeepSeek V4.1 Flash. Any credit pack or plan unlocks all four writers.

The writer you pick handles the whole song: reading your idea, planning the structure, writing every line, checking facts and safety, and rewriting lines that miss. If its service is unavailable, another writer takes over automatically and the studio tells you which one.

Checks that do not rely on AI

Syllable counts and rhymes are also measured by our own rule-based engine, using pronunciation dictionaries for each of the eight languages. Rhyme highlights and length checks come from these rules, not from a model.

Music, images and video

Once your lyrics are ready, these models take over:

Suno v6Developer: Suno
Writes the beat and performs the vocals from your lyrics.
Suno lyric alignment, Whisper (whisper-1)Developer: Suno, OpenAI
Finds when each word is sung, so lyrics stay in sync on screen.
GPT Image 2.5 FlareDeveloper: OpenAI
Paints the scene image for your song while it records.
MiniMax H3 Max, Seedance 2.5Developer: MiniMax, ByteDance
Films the stage performance: MiniMax H3 Max for takes up to 15 seconds (standard 768P, HD 1080P) and Seedance 2.5 for longer takes (480p).
GLM-4V Flash, GLM-5.3 FlashDeveloper: Zhipu AI
Photo check: Zhipu GLM-4V Flash, with GLM-5.3 Flash as backup. Reads the photo once in the dressing room to confirm it shows one adult or one cat or dog and fits our rules.
GPT Image 2.5 FlareDeveloper: OpenAI
Character picture: GPT Image 2.5 Flare turns the checked photo into your illustrated character. It is the same model that draws scene images.

The lyric video preview is assembled in your browser from the scene image, your lyrics and the music. No extra AI model is used for it.

How your words are used

Your words are sent to these models only to create what you asked for, through our API gateway APIMart. We do not use your words or creations to train AI models. Details are in our Privacy Policy.

Photos in the dressing room

A photo you upload is sent to Zhipu AI in China for the photo check and to our image provider to draw your character. It is stored under a random name only until the character is made, then deleted. We keep no face data. The character picture stays in your account until you delete it.

AI-generated content

Everything MakeARap produces, including lyrics, vocals, beats, images and videos, is generated by AI. The voices and performers are original fictional characters, not real artists. Please review your creation before you share it.