Everything we ship to MFL Studio, month by month — new tools, engine upgrades, plan changes and security work.
Faceswaps, character sheets and upscales on one credit balance, a compliance check for ads before they run, two new free tools — and a new studio.
Drop a face and a scene, get that person in that picture — same features, lighting matched to the scene. Three engines to choose from, and if one refuses a photo the run moves to the next on its own. The faces you use are remembered, so the next swap starts where the last one left off.
One photo of a person — or several of the same person — comes back as eight views on one image: the head and the whole body from four sides. It is the input Motion Control needs when all you have is a portrait.
A small, soft or over-compressed image comes back large enough to use again, up to six at a time. Each file is priced on its own pixels before you launch, one line per image, so the expensive one in a drop is never hidden in a total.
Read a creative — or the ad copy alone — against Meta's advertising policies before money goes behind it. You get a verdict, the evidence under it (the exact line, the exact second) and the same creative rewritten so it passes, as one report you can print or send. Free accounts get two reports a month.
Cut a still to the part that matters, or to the exact shape a placement asks for. You aim on a preview, the cut is made on the original at full resolution. Free, no account needed.
See what a photo or clip says about who shot it, where and on what — GPS position, device, date, serial numbers — before anything is removed, then strip it without re-encoding a single pixel. The result is re-read afterwards, so "cleaned" is checked rather than claimed. Free, no account needed.
Video generation, faceswaps, upscales, watermark cleaning and Vision OCR now draw on one monthly balance instead of five separate allowances, so you can spend it all on the tool you actually use. Every price is shown in credits before you launch.
A new look, and the tools grouped into studios. Your settings and dropped files are saved as a draft while you work, a run can chain into the next step on its own (upscale, then clean, then spoof), everything you started is in My projects, and Ctrl+K opens any tool from anywhere.
The Ad Library now opens on a feed of the offers we follow, refreshed every night across six markets, with creatives kept on our side so they never turn into broken images. Browsing it costs nothing; ads Meta pulled are marked as such.
The studio now lives at studio.mflgrowth.com. Old links keep working, and every share link you already sent still opens.
Erase a visible watermark, caption a whole folder in one go, research TikTok ads next to Meta — and a spoof fingerprint that stopped announcing which encoder made it.
Erase a visible watermark from a video or a still — a platform handle, a stock bug, an end card, an AI generator tag. Box the logo on a preview frame and the free engine blurs it out; the Pro engine rebuilds what it was covering on a GPU, which is what an opaque mark over a busy background needs. Two detectors handle the cases boxes cannot: box ONE copy of a repeated logo and we find the rest, or click on the mark and we trace its exact shape across the frame.
Drop up to 200 clips on the Caption Generator, pick one style, and get one archive back. The file count decides the mode — there is no separate page and no switch to find. The lot runs on its own queue, so a few hundred videos never park in front of your interactive runs.
The Caption Overlay now takes stills as well as clips — jpg, png, webp and iPhone HEIC — on their own or mixed into a batch with videos. Timing controls hide for a still, since every caption is composited at once.
A second source next to Meta: TikTok's own Creative Center, browsable by keyword and by industry. Where Meta lists everything an advertiser runs, TikTok ranks what actually performs. Both come back as the same cards, the same archive and the same API payload.
Paste a list of advertiser pages instead of one, and each gets its own run so a single page cannot eat the whole search budget. Single-ad permalinks now resolve too, instead of coming back empty.
Shrink up to 20 files in one drop, on the web and through the API — the only tool that batches, because the compression level is the one setting that means the same thing for every file.
Two fingerprints survived inside files that claim to be iPhone captures: the muxer stamped its own track handler names, and x264 wrote a ~700-byte banner into the bitstream listing its version, settings and this machine's CPU count — identical in every output we shipped, which is exactly what a correlator looks for. Both are gone.
Asking for N copies used to let two of them land on the same playback speed, or trim the same way — a wasted variant, since clipping is what a video hash actually reads. Speed and trim are now dealt out in separate slices, so every copy differs from every other, not just from the source.
Three columns on a wide display, the handset itself is the dropzone, and the burned caption is now the size the preview showed you. Audio files are accepted for transcription, and there is a new IG/TikTok sticker style.
Crypto top-ups have no auto-renew, so a plan used to simply stop. You now get a heads-up two days out and a last call two hours out — and if you are on a team, the message says how many people go down with it. Free trials warn the same way instead of expiring in silence.
A crypto payment arriving a few cents light — normal wallet rounding — used to sit unconfirmed instead of activating the plan.
The Caption Generator opens to everyone, competitor-ad research arrives, a €12 entry tier lands, and every tool page gets a 30-second explainer.
Automatic speech-to-text with reel-style burn-in — dynamic three-word cues, real Apple colour emojis, a font shelf including SF Pro, and visual pickers for size, position, case and style. Add emojis with one click, or edit the transcript before you burn it.
Search the Facebook Ad Library by advertiser or keyword, browse what your competitors are running right now, and pull every creative down as one archive. Searches are kept in a per-operator history.
Put your VAs on your plan. Pro includes 5 extra seats and OnScale 10; they sign in with their own Telegram and inherit your engine and limits. Storage and import budget pool onto the owner, so there is one subscription, not five.
GPU encode speed and the paid tools at a personal volume, between Free and Pro. Annual billing arrived alongside it on every tier.
Each tool page now opens with a short video of the tool doing its actual job on real footage — and hides it once you have used that tool.
b10.studio is now MFL Studio. Same URLs, same API, same accounts — new name and a new face.
The result panel now leads with a plain answer instead of three numbers, shows which preset produced the batch, and previews the video on hover.
14 more countries in the GPS location picker, grouped by continent, and Mirror moved out of the presets to a per-batch toggle where it belongs.
Variants lost to a retry storm are backfilled, so a batch that asked for ten copies ends with ten.
Invisible-watermark removal, one-click respoofing, per-batch location, and a GPU budget that cannot run away.
Strip the invisible SynthID signal that AI generators embed in their output, on a GPU. Different problem from the Watermark Remover, which erases marks you can see.
Re-run any past batch or single variant with different settings, straight from the gallery — the source files load back into staging instead of firing blind.
Choose which country a batch was “shot” in. Each country carries real city anchors, and the timezone offset is derived from the capture instant, so it is daylight-saving correct.
A faster encode path for paid plans, chosen per run.
A time-boxed upgrade that sits on top of your plan and reverts on its own — nothing is downgraded when it ends.
GPU work is now budgeted in compute seconds rather than output count — a spoof can run ten seconds or fifteen minutes, so seconds are what actually bound the cost. A run that would exceed the budget downgrades instead of failing.
The Tools API is documented on the API Access page, with a one-click button that copies the full reference in a form you can paste straight into an AI assistant.
MFL Studio opens to the public — the full spoof engine, a free toolbelt, paid tiers with self-serve billing, a REST API, and an analytics console.
Try every tool as a guest — no account required. Sign in with Telegram to keep your history, lift the limits, and unlock the account-only features. Your guest session folds into your account on first login.
One uploader for clips and stills. Fan a single source into many unique variants in one batch — each samples its own randomization so no two share a fingerprint. Outputs ship as a ZIP or straight to your Google Drive.
Both pipelines strip every original tag, then stamp a randomized-but-consistent iPhone identity — device, iOS build, lens, US GPS, recent date, and the matching photographic EXIF / QuickTime atoms. The video path also rewrites the codec signature so it reads as a clean H.264/HEVC capture.
Pick your protection level. Lite runs on CPU; Moderate, Advanced and Ultra route to a remote GPU engine for a neural perturbation pass, with an automatic fall back to lite if the GPU is unavailable.
Save your own engine configurations on /engine and reuse them per batch. Your settings are yours alone — tuning a preset no longer touches anyone else.
Paste a reel or TikTok link and we pull the source video for you, then run it through the spoofer like any upload (paid plans).
Five side tools alongside the spoofer: a format Converter, a Risk Analyzer (perceptual-hash distance, SSIM fidelity and audio-fingerprint match), a File Reducer, a Frame Extractor, and the Information guide on how detection works.
Free, Pro and Studio tiers gate storage, copies-per-file, daily output, the API and the GPU engine. Upgrade yourself by card (Stripe subscription) or crypto (NOWPayments) — crypto top-ups stack onto any time you already have.
Pro and Studio plans expose a token-authenticated REST API: POST files, poll the batch, pull the variants, or register a signed webhook to be pinged on completion. Mint and rotate tokens on /api-access.
An admin dashboard over every batch, job and output — KPIs, a per-operator breakdown and a media gallery, with the side-tool runs archived and surfaced on the same filters.
Evasion now uses real PDQ for images and Meta’s vpdq for video, fidelity uses an alignment-robust SSIM, and audio is matched with Chromaprint — so the score reflects what platforms actually measure.
The Converter, File Reducer and Risk Analyzer now process on a queue and poll for their result, so the page stays responsive on big files.
Enforced Content-Security-Policy, content-MIME sniffing on every upload, per-IP throttling on guest-session creation, per-user storage caps, and the queue dashboard locked to admins.