ascii-video

Decode the source video with ffmpeg, sample per-frame luminance and edges, then map brightness values to characters from a density-sorted palette on a fixed grid. Composite pre-rasterized font bitmaps onto a pixel canvas and pipe raw RGB frames back to ffmpeg for H.264 encoding.

Analyze the audio with SciPy FFT to extract 6 frequency bands, RMS energy, spectral centroid, and beat detection per frame. Drive value field generators, hue fields, and beat-triggered shaders like glitch bands and chromatic aberration from those features, then encode to MP4 with the audio muxed.

No GPU is required. Rendering runs on CPU with NumPy-vectorized effect math and parallel workers via concurrent.futures, typically at 100-200ms per frame at 1080p. Hardware detection auto-adjusts resolution and worker count based on available cores and RAM.

ASCII characters are small bright dots on black backgrounds, so linear brightness multipliers clip highlights and leave frames dark. Use adaptive percentile-based tonemapping with gamma correction (default gamma 0.75) instead, and verify canvas mean brightness stays above 8.

Long-running ffmpeg encodes deadlock when stderr is piped because the buffer fills at about 64KB. Redirect stderr to a log file instead of subprocess.PIPE, and keep stdin as the raw RGB frame pipe.

Not all Unicode characters render in every monospace font, and Pillow's textbbox returns wrong heights on macOS. Validate palettes at initialization by rendering each character and dropping blank glyphs, and use font.getmetrics() ascent plus descent for cell height.