How a Voice Becomes a Portrait
You can complete a voice test inside Atelier or enter from a CVoice test result. The system reads parameters from the voice report, including timbre composition, voice texture, voice indices, perceived age, and gender-leaning impressions, then turns those cues into a character, clothing, pose, setting, colors, and distinctive details. The result is a portrait card with a name, a poetic line, and a rarity frame.
Each CVoice report generates only one unique portrait. The public portrait gallery, like rankings, inscriptions, and report details support browsing and interaction, but they do not alter the original voice report.
Voice Canvas & Pro Mode
Voice Canvas reads real-time properties such as microphone level and frequency in the browser. Only when the signal exceeds the ambient-noise threshold do circular strokes with varying positions, sizes, colors, and nebula-like textures appear on the canvas. After submission, the system develops the abstract image into a finished work based on the voice trajectory and audio summary.
Pro Mode does not apply Atelier’s preset visual style. Users provide their own prompts, aspect ratios, and reference images. The first reference image supports mask editing, and references can be previewed, enlarged, and reordered.
Creative Boundaries
Image generation requires server-side processing, so completion time depends on queue conditions and generation quality. Voice Canvas and Pro Mode require human verification before the first creation to reduce automated abuse.
Data & Public Display
Only completed voice portraits appear in the public gallery. Users can browse works, report data, rarity, and public inscriptions, and cast one like or dislike. Do not upload images you do not have the right to use or content containing sensitive personal information.