Upload up to 5 Live Photos — the .MOV / MP4 motion part of each one (or a still photo). Drag the big slider under a photo and it tracks the Live Photo frame by frame as you slide. The moment you stop, that frame freezes automatically — it appears big on the card with its timestamp, and the Save button lights up. Tap Save to download it as a full-resolution JPG. Each of the 5 works on its own — stop on a different moment in every one.
Drop up to 5 Live Photos here, or tap “Add Live Photos” above. MOV · MP4 · M4V · still photos too • 0 / 5 loaded • everything converts on this device, nothing is uploaded
Nothing loaded yet. Once loaded: drag the big slider — it tracks the Live Photo as you slide, and the moment you stop, that frame freezes big on the card, ready to save. Tip: on iPhone, open a Live Photo → Share → Save as Video gives you the .MOV to convert here.
Find the one • Out of 20
Your best selfie is already in the burst. This finds it.
Drop up to 20 photos — or the .MOV part of a Live Photo — and get a ranked winner based on face framing, angle, sharpness, light, and expression. Videos are broken down frame by frame so a blink or blur doesn’t ruin the shot. Then, after you pick one, add only a slight chisel to the jaw and take the grey out of your beard — no generative AI, and nothing else changes.
Tip for iPhone Live Photos: In Photos, open a Live Photo → Share → Save as Video, or export both files from iCloud.com / Mac. Upload the still + its .MOV here together.
20photos per batch — exactly what you asked for
12+frames sampled from each Live Photo / video
6scores per image: Overall, Face, Angle, Light, Sharp, Expression
100%private — analysis runs in your browser
How a winner scoresExample
Face & framing
94
Angle
89
Lighting
92
Sharpness
88
Research check: expression and light beat fine angle tweaks — in the largest crowd-rating datasets, a teeth-visible smile moved scores far more than any camera angle did. This scorer weights them accordingly. Full evidence in the Angle Guide below. Change “Prioritize” to re-rank your way.
📸
Drop 20 photos here, or tap to browse
JPG, PNG, WEBP, HEIC previews where your browser supports them, plus MP4 / MOV for Live Photo motion. Everything stays on this device.
Add photos to start. After analysis, tap any photo for its full breakdown and — for videos — every sampled frame.
★ #1 BEST SELFIE
Winner
—
—/ 100 overall
—
Touch-up is optional and only runs after you choose a photo. For a video / Live Photo winner, it works on the best frame and download grabs that frame as a JPG still.
No photos yet — your ranked gallery will appear here. No account, no server upload. Close the tab and the photos are gone.
Breakdown
🎞️ Frame-by-frame — tap a frame
Each frame was scored separately. The highest-scoring frame is marked ★.
For a Live Photo, this enhances the frame you have selected above.
Step 2 · After you choose
Slight improvements only. Same person, same photo.
Start from the photo you picked. Every edit is a local, deterministic slider — jaw shape, beard colour, and slight angle tweaks. Defaults stay subtle or off; how strong it gets is entirely up to you. Check everything with the big Before / After slider below.
🛡️ No generative AI — only local, deterministic edits. Nothing is ever regenerated or redrawn.
No AI face regeneration, no diffusion, no face swap, nothing is invented. This tool only does traditional edits: nudge jaw pixels inward, sharpen the jaw angle, darken grey beard hairs toward your own darker beard colour, rotate the frame a few degrees to level a tilt, and apply a slight 2D perspective warp. Your eyes, nose, mouth, teeth, forehead, hairline, ears and skin are never reshaped or redrawn. The angle tools are honest 2D warps and rotations — not a true 3D re-pose.
Now enhancing: —
Verify it yourself: drag the slider across the photo — left side is your untouched original, right side is the edit. Flick it back and forth over the eyes, nose and background: those should not move or change at all. Only the jaw line and (if it’s on) the beard colour should differ, slightly. To really see a change: turn on Face zoom, hold “Hold to see original”, or run Flicker compare — and watch the glowing outline that shows exactly which area each slider touches while you drag it.
Now adjusting:Jaw chisel · 25The big preview below updates live on every move — no need to let go of the slider.
BeforeAfter
Jaw line
A nudge inward along the lower jaw / chin line only, inside a feathered mask. It now goes meaningfully stronger than before: at 100 the warp reaches about 11% of your face width (previously capped near 4.5%). The default stays Subtle — drag it up only as far as still looks like you. It cannot change your eyes, nose, mouth, forehead or ears, because those pixels are never touched.
🧔 Beard — remove grey
Checking this photo for grey in the beard…
Opt-in only. Darkens just the grey / white hairs in the beard — below the mouth, along the jaw and chin — back toward your darker beard colour nearby. At high strength it now blends grey almost fully back (up to ~95% at 100, previously capped at ~62%), while each hair keeps its own light and shade so the texture stays — never a flat dye. Skin, lips, teeth, head hair and clothes are outside the feathered beard mask, so they don’t change.
📐 Angle — tilt, jaw angle & slight turn
All optional, all default to off. These are honest 2D tools — a rotation and two small warps — not a true 3D re-pose of your head. Within these caps they read as the same photo taken a touch differently.
Rotates the whole frame −8° to +8° so a tilted head / camera reads level. The frame is zoomed just enough to fill the corners — no empty wedges, nothing redrawn.
A separate, stronger pinch focused right at the jaw corners (the gonial angle, just below the ears) — it sharpens the angle where jaw meets neck, rather than slimming the whole jaw like the chisel above. Feathered, so it fades out before the ears and chin.
A subtle horizontal perspective warp centred on the face: one side eases in, the other opens up, simulating a tiny change in viewing angle — a touch more ¾ one way, a touch more frontal the other. Limit: it’s a flat 2D warp, capped low so features don’t distort — it can suggest a small turn, it can’t truly re-pose you.
← more ¾frontalmore ¾ →
✓ Eyes untouched✓ Nose untouched✓ Mouth & teeth untouched✓ Forehead & hairline untouched✓ Ears untouched✓ Skin texture untouched✓ Background never redrawn✓ Head hair & clothes untouched
No generative AI — only local, deterministic edits. The jaw chisel and jaw-angle edits are traditional liquify-style warps inside feathered masks, and the beard edit is a colour adjustment inside a feathered beard mask (below the mouth, with lips and teeth cut out). The face-turn is a small perspective warp feathered around the face, and the tilt tool simply rotates and re-crops the whole frame — it moves every pixel equally and redraws none. None of these can invent a feature: they only move or recolour pixels that are already in your photo. The beard colour fix runs before the warps, so the re-coloured hairs move with the jaw and stay in place.
Defaults are set low on purpose. If you can’t quite tell at first glance, it’s at the right strength — drag the Before / After slider or hit Reset to check.
Step 3 · Optional
New backgrounds — create 5 new photos
One tap lifts you out of the photo you chose (with whatever touch-ups are dialled in above) and sets you into five new scenes. Honest label: these are background replacements — every pixel of the person is your original photo; only the background is new, drawn right here in the app. No generative AI, no face or body changes, nothing leaves your device. Research note (Angle Guide, rule 11): in controlled ratings the backdrop itself showed no significant effect — what these five fix is a cluttered, dark or bathroom/car background, which is exactly what does score down.
Feather softens the cut-out edge — if a hairline or shoulder looks rough, raise it a little and the set rebuilds. If the cut-out itself looks wrong on your photo, the note below will say so; you can still keep and download any of the five.
Pick a photo above first (winner or any card → Enhance), then create the set. Scenes: Café Bokeh · City Golden Hour · Beach Sunset · Evening Bar · Park Daylight.
—
👤
1. Face & angle
Uses your browser’s FaceDetector when available, plus centering, face size, symmetry and skin-region checks as a fallback. Near-frontal, centered, neither too close nor too tiny scores highest — matching the research band in the Angle Guide below: eye level to ~15° above, lead photo within ~15° of frontal. Distance matters more than degrees: at ~30 cm a modelled nose reads ~30% wider, so step back.
💡
2. Light & sharpness
Measures real pixels: exposure (not blown out, not murky), contrast, color vibrance, and Laplacian sharpness for blur / motion shake. Research agrees light is a top factor — direct flash aged faces ~7 years in OkCupid’s 11.4M-opinion EXIF data, and photos taken just after sunrise / before sunset rated best. That’s why a burst frame 0.3s later can outrank the cover still.
🎞️
3. Live Photos, frame by frame
For .MOV / MP4, the analyzer samples frames across the whole clip, scores each one, and surfaces the best still — perfect for catching the moment before a blink or after a smile settles.
✦
4. Then: touch-up, your strength
Only after you choose a photo: a jaw chisel that now goes up to a strong ~11% of face width if you dial it there, grey removal in the beard up to near-full, plus slight angle tools — level a head tilt, sharpen the jaw angle, or nudge the perspective a touch. All local 2D warps and colour edits. No generative AI, no face regeneration.
Crowd-sourced research · Checked against the data
The Angle Guide — there is no magic angle. There is a proven band.
We dug through large-sample crowd ratings (Photofeeler, OkCupid, Hinge reporting), peer-reviewed camera-angle studies, photographer technique, and community consensus (Reddit review culture, Photofeeler reviewer panels, looksmaxxing “anglemaxxing” threads). No study finds a magic angle like “17° above, 32° turned.” Labs only ever tested 0° vs ±30° — every finer degree band below is photographer/community convention, and it’s labelled that way. What the evidence does converge on: expression beats angle, distance is the most quantified factor, and light can make or break the jaw. Evidence grades: A = large-sample crowd/platform data (10k+ ratings) · B = peer-reviewed study · C = community consensus · D = creator/photographer opinion.
1.2–1.8 mShoot from 4–6 ft away — tripod + timer or someone else holding the phone. At ~30 cm, a modelled nose looks ~30% wider (Ward et al. 2018, JAMA Facial Plastic Surgery). Source
0° to +15°Camera height band: eye level is the measured trust optimum (5.03/9 vs 4.27 from below); a slight raise is the consensus jaw/slimming sweet spot. Penalise anything below eye level. Source
+2.55 trustOpen-mouth smile vs no smile for men’s dating photos (Photofeeler, “millions” of photos — vendor data, correlational by its own caveat). 54% of men’s dating photos had no smile at all. Source
Flash ≈ +7 yrsDirect on-camera flash cost the equivalent of ~7 years of attractiveness in OkCupid’s EXIF analysis of 11.4M opinions / 552,000 photos. Best-rated photos: just after sunrise, just before sunset. Source
📐 Camera height — vertical angle
Evidence B + C/D
Do this: lens at eye level to ~15° above. Acceptable to ~+20°. Never below the chin for a lead photo — heavy penalty past −10°.
Trust peaks at exactly eye level (Baranowski & Hecht 2018: 5.03/9 at eye level vs 4.57 above, 4.27 below — and elevation alone did not change attractiveness). A raise only helped attractiveness combined with a 30° side turn (Schneider & Carbon 2017). Men-specific warning (Makhanova et al. 2017, Psychological Science): from-below made men look more dominant — mostly to other men — not more attractive. Too high (≳30°): forehead enlarges, jaw recedes, you read younger/smaller. Too low: up-the-nose, under-chin emphasis, face judged heavier. The “hold it 10–15° above” figure itself is photographer/community convention, not a measured optimum.
Do this: lead photo — face within ~15° of frontal so gaze and smile read. Posed/secondary shots — body turned 30–45°, face turned only ~15–30° back to the lens. Full profile (90°): never for a lead.
A modest turn beat dead-frontal for attractiveness in lab ratings (tested at 30°; effect stronger for female faces). There is no winning side: the direct left-vs-right comparison was null. The famous “left cheek” finding (Lindell 2017: 41% of 2,000 Instagram selfies showed the left cheek vs 31.5% right) is a posing preference — people choose their left — not evidence raters score it higher. So: don’t chase a side; test both and keep the one you consistently prefer.
Do this: 1.2–1.8 m (4–6 ft). Minimum ~1 m. Use the 2× lens or crop afterwards — never solve framing by moving closer.
Ward et al. 2018 modelled it: at 12 in (30 cm) the nasal base appears +30% wider in men (+29% women), tip +7%; at 5 ft, essentially no distortion. Caveat: that’s a geometric model, not photos of real people. The rating cost is real, though — Bryan, Perona & Adolphs 2012 (Caltech) photographed the same faces close (0.45 m) vs far (1.35 m) and close faces got lower trust, competence and attractiveness ratings, and viewers couldn’t say why. It’s the distance that distorts, not the lens — a longer lens only helps because it forces you back.
Do this: push your forehead toward the camera (neck extended), then drop the chin just ~5–10°. Never tuck the chin back into your neck; never jut it up.
This is Peter Hurley’s core move: “Jam your forehead towards the camera and allow the chin to follow, tightening the skin against your jawline.” It’s unanimous photographer technique (grade D), and the extremes are study-backed: pitching the face up raises perceived dominance but lowers attractiveness (Marshall et al. 2020), and a bowed head reads submissive/sad. Forward-then-slightly-down is specifically not a bow — the extended neck is what stops an under-chin fold forming.
Do this: lead photo — eyes into the lens and a genuine smile, teeth visible. Add a slight “squinch” (lower lids raised a touch) rather than wide-open eyes.
Photofeeler’s 60,000+-rating study: teeth-visible smile = +1.35 likability — “by far the most impactful characteristic we found” — plus +0.33 competence; squinch added +0.33/+0.22/+0.37. Direct gaze is the trust winner in lab work, and direct gaze + smile is the classic attractiveness combo. Averted gaze is fine for candid secondary shots (downward-averted reads best of those; sideways/upward worst). Sunglasses in the lead photo: −0.36 likability — eyes must be visible.
Do this: soft, directional light — golden hour outdoors, or face a big window indoors from ~3–4 ft, angled ~45° to it. Light slightly above eye level so a soft shadow forms under the jaw.
Direct flash is the villain (OkCupid EXIF: ≈ +7 years; most flash photos in the data were direct, not bounced). Overhead indoor downlights — the bathroom/office look — shadow the eyes, nose and chin and age the face; too-dark photos cost −0.38 likability (Photofeeler). Overcast sky is a free softbox. Ring lights are a divided creator opinion with no rating data: acceptable fallback in a dark room at eye level or slightly above, but window/golden-hour light beats them — and reads more natural.
Recipe: camera at/slightly above eye level + forehead forward, chin down a touch + slight ¾ turn + side light so one side of the jaw falls into shadow + distance ≥1.2 m.
The measurable part is the shadow line: a shadow outlining the jaw all the way around scored Competence +0.24, Likability +0.18, Influence +0.18 in Photofeeler’s 60k-rating study. Jaw definition is a lighting outcome as much as an anatomy one — taut skin (neck extended) is what gives light an edge to cut. Biting your back teeth together for the shot is a real photographer trick to flex the jaw muscle for one instant — no study behind it, and habitual clenching risks jaw strain, so treat it as a 1-second pose at most.
Do this: lead photo = bust (head-and-shoulders) or torso framing, clean, bright, uncluttered, background softly blurred (Portrait Mode). Avoid: bathroom mirror, car interior, messy room, gym mirror.
The honest catch: in Photofeeler’s controlled business study, the setting itself — studio vs nature vs urban vs home — had no significant effect at all. What dating data rewards is the activity/social signal, not the backdrop: Hinge reporting (secondary, no primary report retrievable) put sports at +75% likes, night-out/social +74%, candid +15% vs posed. Framing is solid A-grade: face-only close-ups −0.21 likability; full-body as a lead −0.29 competence/influence (a full-body shot still belongs later in a set). Shallow depth of field rated best in OkCupid’s EXIF data. Heavy filters: Hinge reporting put Snapchat-filter photos at −90%.
Camera height: eye level to ~15° above scores highest (0° = lab trust optimum; slight raise = consensus). Below eye level loses progressively; past −10° and past +30° lose heavily.
Turn, don’t face dead-on — but keep the lead near-frontal: body 30–45° with face 15–30° back for posed shots (lab: 30° lateral > frontal); lead photo face within ~15°. No side is scored — left and right tested equal.
Distance ≥1.2 m or it isn’t a fair photo: arm’s-length ≈ +30% modelled nose width plus lower trust/attractiveness ratings. Tripod + timer, 2× lens, crop later.
Chin rule: forehead forward, then chin down a touch (~5–10°). Chin tucked back (double-chin fold) and chin jutted up (dominance up, attractiveness down) both score down.
Gaze rule for the main photo: eyes into the lens. Averted gaze is neutral for candids, negative for a lead.
The one forbidden combination: eye contact + no smile. (Photofeeler replication, n=7,140.) Eyes on the lens means a smile is required for a top expression score.
Smile with teeth = top expression tier (men: Trust +2.55 open / +1.86 closed vs none). Closed-mouth smile = second tier. Serious/neutral = bottom tier for a lead, fine for candids. Slight squinch bonus; no wide-eyed stares.
Eyes must be visible in the lead photo: sunglasses, hat shadow or hair over the eyes = hard penalty (sunglasses −0.36 likability in 60k ratings).
Light: soft daylight wins; direct flash, harsh midday, backlight and too-dark frames lose. Side light at ~45° leaving a soft jaw shadow = jaw-definition bonus (+0.24/+0.18/+0.18).
Jaw shot recipe (pose + light, not anatomy): slightly-above camera + chin forward/down + ¾ turn + jaw side in shadow.
Background: clean, bright, uncluttered, softly blurred; the real bonus is activity/social context (sports +75%, night-out +74% — Hinge, secondary reporting). Hard penalties: bathroom, car, messy room, gym mirror. Stated honestly: the backdrop itself showed no significant effect in controlled ratings — clutter, darkness and context are what score.
Framing: bust or torso for the lead (face-only −0.21 likability; lead full-body −0.29). Oversaturation and heavy beauty filters score down (filters −90% likes in Hinge reporting).
🚩 Myths & traps — checked, and not built into this app
“Men: look away and don’t smile” (OkCupid 2010) — failed replication. Photofeeler re-ran it matched-for-demographics on 7,140 men’s photos (2017): the result did not reproduce. The only harmful combo was eye contact without a smile. The original study had also removed the most/least attractive 40% of photos before analysing — the likely flaw. Replication
Mewing does not reshape an adult jaw. No clinical trials support it; the American Association of Orthodontists warned in Jan 2024 that it oversimplifies facial structure, with risks (loosened teeth, bite shift) outweighing unproven benefits. Before/afters are growth, weight change, posture, lighting and angle — the photographic tricks above. It is not in this app’s tips as a method. Mewing evidence overview
“Your left side is your best side” — not a rating win. People pose their left cheek more (41% of selfies); raters did not score it higher when left and right were compared directly.
Men-specific Hinge numbers (+102% forward-facing, −40% selfies…) are secondary-only. They’re everywhere in guides but no primary Hinge report was retrievable — directionally consistent with everything else, quoted here only as secondary reporting.
There is no Tinder angle study. No comparable public photo-factor study from Tinder was found; “Tinder study says X%” claims in SEO guides trace back to Hinge/Photofeeler/OkCupid — or to nothing. Two widely-cited “Photofeeler 2023/2024 studies” (men overrating their photos by 2.3 points; background as 3rd-biggest factor) could not be found on Photofeeler’s blog at all and are treated as likely fabricated citations.
Open questions, stated plainly: no study tests the exact degree bands apps want (labs used 0° vs ±30° only), so all finer bands here are convention. Photofeeler’s 2026 smile analysis publishes no exact sample size (“millions”) and is correlational vendor research, not peer-reviewed; its 2014 study is business-context, not dating. Community consensus is grade C — reconstructed from a Reddit mirror, Photofeeler’s published reviewer panels, and independent review-culture summaries that agree with each other, not from a coded sample of threads. And the men’s rules don’t transfer to women unchanged: from-above does raise women’s rated attractiveness, and the turn effect was stronger for female faces.
Honest limits: This is a web-first version running entirely in the browser. It cannot directly unpack Apple’s paired HEIC+MOV Live Photo container as one file in most browsers, and facial-expression reading is an estimate from pixel cues (eye-region contrast, sharpness, symmetry), not a medical or biometric ID system. It never identifies who a person is. For the sharpest results, upload the original files — screenshots and compressed re-saves score lower on sharpness, correctly.