Video Washer
Drop videos for a fresh server-side wash — motion, shading, metadata strip, and randomized re-encode, plus an optional IG filter — delivered back as a single zip.
Wash types 0 how the camera moves
Pick none if you only want a clean spoof — full metadata strip + fake EXIF / encoder + invisible noise + re-encode still runs (one output per file). Use Metadata only in the chip grid for the same effect explicitly.
Each wash chip also has an IG filter on / off toggle at the bottom. Default is on (that wash type pulls a filter from the IG filter set below). Click it off to keep that specific wash type's outputs filter-free.
IG-style filters 0 optional, multi-select
Filters are cosmetic stickers layered on top of the wash, not a replacement for it. The metadata strip + fake EXIF + noise + re-encode all still happen no matter what you pick here.
Each output randomly picks one filter from your selection. Pick "No filter" to mix in clean outputs alongside the styled ones.
Drag the bar at the bottom of each chip to dial that filter's strength up or down. Double-click the bar to snap back to default.
Picture / video quality
Default runs full-strength wash for maximum variation between outputs.
Image-only extras (videos always get fresh randomization automatically)
Image Washer
Edit photos per image (blur, caption, crop, rotate), then wash each with a randomized brightness, contrast, saturation, blur, noise, hue, temperature and gamma pass.
IG-style filters 0 optional — leave empty for "wash, no filters"
Drag the bar at the bottom of each chip to dial that filter's strength up or down. Double-click the bar to snap back to default.
Output settings
Wash Settings
Pixel Spoofer
Calibrated re-encode that defeats AI image classifiers — drop AI-generated images, hit Spoof, and get back files that read as non-AI. 2 credits per image; output is always .jpg.
Spoofed batch
Image Generation
Prompt an image, pick a character, preview the credit cost, and generate one output or a batch.
Output
Your latest image or batch appears here. Saved production history stays on the History page.
Result
Batch 0 / 0
History 0
Train Character
Upload reference images, review captions, and train a reusable character model you can use across image, video, and Hive Flow workflows.
JOYCAPTION_RUNPOD_ENDPOINT_ID and
LORA_TRAINING_RUNPOD_ENDPOINT_ID on the server
to enable live training.
Step 1 of 4 — Upload your reference images
Character
The trigger word is added to every caption used for training,
and is the unique token you'll use when generating with
this character later. Pick a token that doesn't collide with
everyday English (e.g. luxdek0tal0, not
portrait).
Reference images
.txt caption files (same basename as the image).
0 / 50
Click Auto-caption and we'll generate a description for each image. The trigger word from Step 1 is automatically added to the start of each caption — what you see here is exactly what training will use.
—
portrait of <trigger> in a forest.
Loading pricing…
Worker log tail (last lines)
Diagnostics (paths, counts)
Sample previews by checkpoint step
Your trained characters
Workflow
Workflow
Add up to 16 reference images. Reference media is sent to our generation provider to create your output.
Your subject copies the movement from this clip — the reference sets the motion, not the identity.
Your subject copies the movement from this video — it sets the motion, not the identity. 3–30 s, up to 100 MB. MP4 / MOV / MKV.
Generating…
Result
History
Subscription
Loading…
Top up
Top-up credits never expire. Subscription required.
