Reel Max
9.0The best clip in the test: a natural stride, hair and shirt moving with her, and the last frame is still unmistakably Melanie. Comes with generated street sound.
Every Flashlook agent got the same model, the same three-piece look and the same pinned brief. We reviewed every frame against the originals and ranked them on what a catalog actually needs: the right face, the right clothes, a real photograph.
What every agent received


Brief 1
Studio close-up
Tests: Face likeness, skin, fabric texture and how the layers sit.
Brief 2
Street, full body
Tests: Proportions, the jeans' silhouette, the scene and the light.
Three rankings, three winners. Select a card to open its full review.
Our pick for each kind of shoot. Select a card to open its full review.
Switch between the overall, quality and cost-effective rankings. Open any row for both frames, the six scores and our notes.
The all-round pick: 60% image quality, 40% cost-effectiveness.
The studio default for good reason - recognisably Melanie, every garment right, natural light in both briefs.
1632×2448 output · 4 MP · median render 2m 20s
The same face crop from every tier of a family. The frame stays the same - the detail you can crop into grows.
Precision-tuned. The deepest detail at the top of the ladder.
Same crop of the studio close-up for every tier. Composition, likeness and garments stay locked across the ladder; what grows with each tier is how much skin, freckle and hair detail survives a close crop.
Both briefs and every score, with the winner of each row in bold.
It keeps the model's exact face and every garment detail at catalog size, for 80 credits - less than the Optic and Flare everyday tiers that match it. That balance is why the studio starts on it.
Every agent got the outfit roughly right. Only the Optic, Flare and Lumen families kept Melanie unmistakably Melanie in both briefs - the narrow face, high cheekbones and freckles. The others drifted toward a generic model.
Within Optic, Flare and Lumen the composition and likeness stay locked while skin, hair and fabric detail grow with each tier. Medium is right for product grids; High and above hold up for hero images, close crops and print.
Bolt 4K returns 3392×5056 (17 MP); everything else lands between 2.8 and 4.3 MP. Use it when the pixel count matters more than the face.
Halo followed the brief with natural skin. Glint and Ember ignored the close-up, moved the street shot into a gray studio and changed the garments. Start on a quality agent and switch only when a category is refused.
Model likeness
Is it unmistakably Melanie - face shape, eyes, freckles, hairline, smile?
Photorealism
Would it pass as a real photograph - skin texture, hair, hands, light falloff?
Garment accuracy
Mineral wash on the tee, rolled sleeves, the shirt worn open, the balloon leg - nothing invented, nothing dropped.
Follows the brief
Framing, pose, backdrop, location and light exactly as requested.
Image craft
Composition, light and atmosphere - does it look like a considered editorial frame?
Fine detail
Does it hold up at close crops - pores, freckles, individual hairs, weave and wash in the fabric?
Each video agent animated the same street frame with the same prompt. Clips play silently on loop - the sound badge shows which agents add audio.
Same start frame · Lumen Refined street shot
“She keeps walking toward the camera along the city sidewalk at an easy pace, the open light-blue shirt swinging with her stride and the balloon-leg jeans moving naturally, then glances at the lens and smiles. Steady tracking camera, soft cloudy daylight, natural street ambience with distant traffic and footsteps.”
4-second clips · sound requested on every agent that offers it
The best clip in the test: a natural stride, hair and shirt moving with her, and the last frame is still unmistakably Melanie. Comes with generated street sound.
Walks toward the lens, glances and smiles exactly as asked, with a sharp, faithful face at the end. A touch less fluid than Reel Max.
Does everything the prompt asks - walks in, glances at the lens, smiles - and keeps the balloon jeans and open shirt. The face softens slightly and the freckles fade, but it is still Melanie. Comes with street sound, in about a minute.
Follows the prompt and keeps the outfit, but the face softens and drifts by the final smile.
Rendered in 15 seconds for 71 credits. Simple motion and a slightly wider smile, but the look holds - ideal for quick social loops.
Fast and permissive, but by the last frame it is a different woman in sunnier light, and the balloon jeans have straightened out.
This agent's own engine declined Melanie's frame, so the clip you would have received came from a backup. Shown for transparency, not ranked.
The most dynamic movement of the set, but the face rounds out by the end.
Flow's own engine declined this face as a possible real person on every run, so our safety net rendered the clip with Reel's engine. This is what you get on that refusal - not Flow's native output.
Motion
Natural gait, weight and camera movement - no warping, sliding feet or rubber limbs.
Identity hold
Is it still Melanie in the last frame?
Garment hold
The open shirt, tee and balloon jeans stay the same garments throughout.
Follows the prompt
Walks toward the camera, glances at the lens, smiles.
Flashlook agent benchmark (November 2026). 22 photo and 7 video agents, one model, one look, one brief. https://flashlook.ai/benchmark/agents
Every agent is available on every plan, including free. Pick one in the studio and see how it handles your garments.