{ "slug": "mel-spectrogram-generator", "name": "gizza-ai/mel-spectrogram-generator", "version": "0.1.0", "title": "Mel Spectrogram Generator — gizza.ai", "description": "Render a local audio file as a PNG mel spectrogram with ML-style FFT, hop, mel-band, dB, colormap, and image-size controls.", "tags": [ "audio", "spectrogram", "mel", "machine-learning", "png", "features" ], "category": "audio", "urls": { "page": "https://gizza.ai/tools/mel-spectrogram-generator/", "markdown": "https://gizza.ai/tools/mel-spectrogram-generator/index.md", "descriptor": "https://gizza.ai/tools/mel-spectrogram-generator/tool.json", "deep_link_example": "https://gizza.ai/tools/mel-spectrogram-generator/?url=https://example.com/input&input_format=base64&n_fft=2048&hop_length=0&n_mels=128&fmin=0&fmax=0&mel_scale=slaney&window=hann&scale=db&top_db=80&db_reference=peak&colormap=magma&width=1000&height=400&channel=mix¢er=true&resample_hz=0" }, "cli": "gizza tool mel-spectrogram-generator 'url=https://example.com/input' 'input_format=base64' 'n_fft=2048' 'hop_length=0' 'n_mels=128' 'fmin=0' 'fmax=0' 'mel_scale=slaney' 'window=hann' 'scale=db' 'top_db=80' 'db_reference=peak' 'colormap=magma' 'width=1000' 'height=400' 'channel=mix' 'center=true' 'resample_hz=0'", "tool": { "description": "Render an uploaded or pasted audio file as a mel-scaled PNG spectrogram. The tool decodes common audio containers locally, downmixes or selects a channel, optionally resamples, applies an STFT, triangular mel filterbank, dB/power scaling and a selectable colormap. It returns an image envelope plus a short report with the resolved FFT, hop, sample rate, mel bands, peak band and PNG size.", "parameters": { "additionalProperties": false, "properties": { "center": { "default": true, "description": "Reflect-pad by n_fft/2 before framing, matching librosa center=True (default). Turn off for only complete in-file frames.", "type": "boolean" }, "channel": { "default": "mix", "description": "Audio channel selection before analysis: mix (default downmix to mono), left, or right. Mono sources fall back to their only channel.", "enum": [ "mix", "left", "right" ], "type": "string" }, "colormap": { "default": "magma", "description": "Color palette for the PNG: magma (default), inferno, viridis, plasma, turbo, grayscale, or grayscale_inverted.", "enum": [ "magma", "inferno", "viridis", "plasma", "turbo", "grayscale", "grayscale_inverted" ], "type": "string" }, "db_reference": { "default": "peak", "description": "dB reference for scale=db: peak maps the clip's loudest cell to the top color (default); full_scale keeps quiet clips visibly dimmer.", "enum": [ "peak", "full_scale" ], "type": "string" }, "fmax": { "default": 0.0, "description": "Highest mel filter edge in Hz (0-24000, default 0 = Nyquist for the analysis sample rate). Values above Nyquist are clamped.", "maximum": 24000, "minimum": 0, "type": "number" }, "fmin": { "default": 0.0, "description": "Lowest mel filter edge in Hz (0-24000, default 0). Raise it to ignore rumble or telephone-band low cut.", "maximum": 24000, "minimum": 0, "type": "number" }, "height": { "default": 400, "description": "Output height in pixels (0 = one row per mel band, otherwise 32-2048; default 400).", "maximum": 2048, "minimum": 0, "type": "integer" }, "hop_length": { "default": 0, "description": "Samples between adjacent STFT frames (0 = n_fft/4, default 512 when n_fft is 2048). Smaller hops make a wider, denser image.", "maximum": 8192, "minimum": 0, "type": "integer" }, "input": { "description": "Audio file bytes encoded as base64 or hex. Decodes WAV, AIFF, CAF, FLAC, MP3, OGG/Vorbis, MP4/M4A, MKV/WebM and AAC-ADTS; the selected channel is rendered as a PNG mel spectrogram. Limit 24 MiB input bytes and 4,000,000 decoded samples.", "type": "string" }, "input_format": { "default": "base64", "description": "Encoding used for the pasted audio bytes: base64 (default) or hex. The standalone page fills this from the uploaded file automatically.", "enum": [ "base64", "hex" ], "type": "string" }, "mel_scale": { "default": "slaney", "description": "Mel frequency formula: slaney (default, librosa-style area-normalized filters) or htk (classic 2595*log10 scale).", "enum": [ "slaney", "htk" ], "type": "string" }, "n_fft": { "default": 2048, "description": "FFT and analysis-window size in samples (64-8192, default 2048). Larger windows sharpen frequency detail but blur time detail.", "maximum": 8192, "minimum": 64, "type": "integer" }, "n_mels": { "default": 128, "description": "Number of triangular mel bands before image scaling (8-512, default 128). Common ML front ends use 64, 80 or 128 bands.", "maximum": 512, "minimum": 8, "type": "integer" }, "resample_hz": { "default": 0, "description": "Resample before analysis (0 = keep the source rate, otherwise 4000-48000 Hz). Use 16000 or 22050 to match common ML pipelines.", "maximum": 48000, "minimum": 0, "type": "integer" }, "scale": { "default": "db", "description": "Pixel scaling: db (default log/dB spectrogram), power (linear mel power), or magnitude (square-root power).", "enum": [ "db", "power", "magnitude" ], "type": "string" }, "top_db": { "default": 80.0, "description": "Dynamic range shown below the reference level when scale=db (10-160 dB, default 80). Lower values increase contrast.", "maximum": 160, "minimum": 10, "type": "number" }, "width": { "default": 1000, "description": "Output width in pixels (0 = one pixel per STFT frame, otherwise 32-4096; default 1000).", "maximum": 4096, "minimum": 0, "type": "integer" }, "window": { "default": "hann", "description": "Window applied before each FFT: hann (default), hamming, blackman, or rectangular/no taper.", "enum": [ "hann", "hamming", "blackman", "rectangular" ], "type": "string" } }, "required": [ "input" ], "type": "object" } } }