GAVEN ECHOES

Build Temporal Echo loops, recursive zooms, dream drifts, and animated exports.

MINI MAGIC · CREATURE LAB

Choose a photo. Press one button.

A lightweight browser DeepDream experience using the stable Inception engine underneath.

STABLE

CHOOSE PHOTO

Magic strength Output

Preset Texture scale Seed lock Original ↔ Dream 100% dream

Dream brain Inception ready Passes MobileNet intensity Preset stack

PASS HISTORY

No passes yet.

STYLE TRANSFER · ARBITRARY NEURAL STYLE

Choose a separate style image and apply its visual style to the current content image entirely in the browser.

CHOOSE STYLE IMAGE

No style image selected.

Style strength80% Processing size Content source After transfer

STYLE TRANSFER RESULT

Models load only when style transfer is requested.

AUDIO REACTIVE · MUSIC-DRIVEN VISUALS

Choose an audio file from this device. The audio stays local in the browser and drives the visual engine in real time.

CHOOSE AUDIO FILE

No audio selected. Audio stays on this device.

Visual preset Sensitivity 100% Smoothing 70% Visual intensity 65%

LIVE AUDIO DREAM

AUDIO SPECTRUM

BASS0

MIDS0

TREBLE0

TRANSIENT0

AUDIO MAPPING · ADVANCED

Bass controls Mids controls Treble controls Transient action

Choose an image and an audio file to enable Audio Dream.

MUSIC STRUCTURE ANALYSIS · AUTO DIRECTOR

Analyze the selected track locally, detect energy changes, then build a music-aware Director timeline.

Directing style Scene density Minimum scene length Analysis resolution

PEAK ENERGY0

CHANGE POINTS0

TRANSIENT EVENTS0

SECTIONS0

Choose an audio file first.

AUDIO VISUAL DIRECTOR · SCENE TIMELINE

Build timed visual scenes for the song, then play or render the full sequence.

Default transition Auto scene length

NOWNo scene

Choose an image and audio file, then add scenes.

AUDIO RENDER CAPTURE · VIDEO EXPORT

Record the live Audio Dream canvas together with the selected song. Recording happens locally in the browser.

Capture resolution Capture frame rate Video quality Include audio

Choose an image and audio file first.

00:00.0 · 0 frames

FEEDBACK ANIMATION · LIVE LOOP LAB

Generate fast non-neural animation frames from recursive feedback or evolving reaction diffusion. Designed to be lighter than neural Temporal Echo.

Animation mode Frames Frame size Frame delay Motion amount 45% Feedback amount 72% Reaction preset Sharpen between frames Source reinjection 10% Edge reinforcement 35% Feedback decay 88% Resample quality Detail floor 20% Loop preview

Choose a photo to enable Feedback Animation.

0 / 0

No frames yet.

TEMPORAL ECHO · ANIMATION LAB

Generate related dream frames one at a time, preview the loop, then export an animated GIF or PNG frame ZIP.

Animation mode Frames Frame size Frame delay Motion amount 42% Dream strength Loop preview

Choose a photo to enable Temporal Echo.

0 / 0

No frames yet.

CHAIN COMPOSER · AUTOMATION STUDIO

Build a sequence of stages, reorder them, and run the whole chain with one click.

Batch variations Seed mode Auto-save each result

Choose a photo to enable Chain Composer.

BATCH RESULTS

No batch results yet.

ATTENTION LAB · EXPERIMENTAL

A lightweight Vision Transformer estimates which image patches matter to its interpretation. The relevance map can then bend or echo the displayed dream.

Map source Patch grid Attention effect Effect strength 45%

TRANSFORMER / PATCH RELEVANCE

ViT loads only when requested. Local fallback is always available.

VISION LAB

Optional guide models shape where the finished dream appears. They do not replace the stable Inception/MobileNet dream engines.

Edge protection Edge amount 45% Subject behavior Depth behavior Guide strength 70%

SUBJECT MASK

DEPTH MAP

EDGE MAP

Edge and Fast Subject guides run locally with no model download. AI Subject and Depth are optional.

FOCUS MASK

Paint areas where the dream should stay strongest. Unpainted areas can preserve more of the original photo.

Brush size Outside dream 20% Feather 8 px

Drag on the mask canvas to paint dream-focus areas.

Choose a photo to begin.

Legacy v0.17 presets restore their original region, channel mode/count, steps, strength, octaves, scale, recursion, mutation, and seed.

ORIGINAL

DREAM 0

ADVANCED CONTROLS

CHOOSE FROM GALLERY TAKE PHOTO

Layer region Channel mode Channels / layer 12 Seed Steps / octave 14 Strength 0.85 Octaves 3 Octave scale 1.18 Random shift 8 Working canvas Recursive growth 0.55 Feature strategy Channel A 32 Channel B 96 Channel C 160 Recipe A 0.60 Recipe B 0.25 Recipe C 0.15 Channel drift 0.12 Feedback amount 0.30 Feedback zoom 1.08 Feedback rotation 0 Feedback blend Brightness preserve 0.85 Contrast preserve 0.70 Persist recipe Dream direction Spatial frequency Freeze A Freeze B Freeze C Octave recipe Detail recursion 0.35 Recipe name Memory mode Pass cap Layer mutation 0.18

Downloading Inception V3…

HALLUCINATION HISTORY

No dreams yet.

PRESET LIBRARY

SESSION GALLERY

PROJECT SESSIONS