AI Music Generator

Accepts user-provided lyrics or text descriptions and generates complete original songs by encoding input text through a neural composition model, then conditioning generation on discrete style parameters (genre, mood, tempo, instruments, vocal gender). The system processes parameterized requests through a cloud-based inference pipeline and outputs multi-format audio (MP3, WAV, MIDI) within claimed <1 minute latency. Generation is queued based on tier-dependent concurrency limits (1 for Free/Basic, 10 for Standard, unlimited for Pro).

Generates original song lyrics from user-provided semantic inputs (theme, keywords, genre, emotion, duration, language, song structure) using a text generation model conditioned on these discrete parameters. The system accepts structured input (theme up to 1000 chars, keywords up to 300 chars) and outputs formatted lyrics with specified verse/chorus structure. This capability is decoupled from music generation, allowing users to generate lyrics-only or use generated lyrics as input to the music generation pipeline.

Implements a credit system that limits daily music generation volume based on subscription tier. Free tier users receive 20 credits/day (approximately 4 songs/day at 5 credits per song inferred). Paid tiers offer higher daily quotas (Basic ~33 songs/month, Standard ~167 songs/month, Pro ~400 songs/month). Credits reset daily and appear to roll over if unused (based on pricing language 'unused credits roll over'). This mechanism enforces fair resource allocation and creates upgrade incentive for high-volume users.

Conditions music generation on discrete categorical style parameters (genre, mood/vibes, tempo, instruments, vocal gender) selected from predefined dropdowns and multi-select lists. The generation model uses these parameters as conditioning signals to shape the output music characteristics. Users can also specify 'Random' for any parameter to allow the model to choose. This parameterized approach enables rapid style variation without changing lyrical content.

Allows users to specify styles, genres, or characteristics to EXCLUDE from music generation through an 'Exclude styles' parameter. This negative prompting approach enables users to specify what they don't want in the output, complementing positive style conditioning. Implementation details (how exclusions are encoded and enforced) unknown.

Processes user-uploaded audio files through a source separation model that isolates and removes vocal tracks, outputting a clean instrumental version. The system accepts audio uploads (WAV/MP3 format inferred) with tier-dependent duration limits (1 min free, 2 min Basic, 8 min Standard/Pro) and applies neural source separation to decompose the audio into vocal and instrumental components. Output is provided in the same formats as music generation (MP3, WAV, MIDI for paid tiers).

Generates cover versions of songs by applying user-selected or custom voice models to existing song audio or lyrics. The system accepts audio uploads or text input and synthesizes vocal performances using neural voice conversion or text-to-speech models conditioned on voice parameters (gender, custom voice model). Generated covers are output in standard audio formats and can be downloaded or shared. Implementation details (whether voice conversion or TTS-based) are unknown.

Trains personalized voice models from user-provided audio samples, enabling voice synthesis and cover generation using the trained model. The system accepts audio uploads (format unknown) and trains a neural voice encoder/decoder model on the provided samples. Trained models are stored in the user's account and can be applied to music generation, cover generation, and singing photo features. Training capacity is tier-dependent (100 models max for Basic, unlimited for Standard/Pro).

Generates singing videos by synthesizing lip-sync and facial animation from a static photo input, combined with generated or uploaded audio. The system accepts image files (format unknown) and audio input, then applies neural video synthesis to create a video of the person in the photo singing. Output duration is tier-dependent (10 seconds free, 2 minutes Basic, 10 minutes Standard/Pro). Generated videos are downloadable and shareable.

Exports generated or processed audio in multiple formats (MP3, WAV, MIDI) with format availability determined by subscription tier. Free tier users can only export MP3; paid tiers unlock WAV (lossless audio) and MIDI (symbolic music notation for DAW import). Export is performed after generation completes, with files available for immediate download. MIDI export enables integration with external DAWs for post-processing.

Stores generated songs and projects in cloud storage with retention duration determined by subscription tier (30 days free, 365 days Basic, unlimited Standard/Pro). Users can access, download, and manage stored projects through the web interface. Storage is tied to user account and persists across sessions. Free tier projects are automatically deleted after 30 days; paid tiers offer longer or indefinite retention.

Manages simultaneous music generation requests through a queue system with tier-dependent concurrency limits (1 for Free/Basic, 10 for Standard, unlimited for Pro). When users exceed their concurrent limit, new generation requests are queued and processed sequentially. Queue position and estimated wait time are unknown. This mechanism enforces resource allocation and creates upgrade incentive for users with high generation volume.

Manages intellectual property rights for generated music through subscription tier. Free tier users cannot use generated music commercially and must credit 'AI Song Maker' in any public use. Paid tier users (Basic, Standard, Pro) receive commercial licensing rights, allowing monetization on platforms like YouTube, Spotify, and other distribution channels. Rights persist after subscription cancellation for songs generated during active subscription. Licensing terms and exclusivity unknown.