Black Forest Labs
FLUX 3
At a glance
Strengths
- Perfect 5.00 on on-screen text: "OPEN 24H", "FRESH BREAD DAILY" and "SLOP" on a moving t-shirt all stay legible in every frame.
- Perfect 5.00 on lip-sync: all four spoken and sung lines are word for word in the audio, checked by transcription.
- Native audio at no extra cost on Black Forest Labs' own API, and clips up to 20 seconds.
Watch-outs
- Hands are its weakest category (4.38): pepper slices multiply under the knife and the shoelace bow is never finished.
- The 360-degree orbit stops at roughly 90 to 120 degrees, and the ice cube barely melts.
When to use it: Anything with readable text, dialogue or a sequence of cuts that must keep the same character and place.
Capabilities
| Native audio | Yes |
| Multi-shot sequences | No |
| Image to video | Yes |
| Public API | Yes |
| Max resolution | 1080p |
| Max duration | 20s |
Benchmark scores
| Overall | 4.83 / 5 |
| Prompt fidelity | 4.80 |
| Visual quality | 4.94 |
| Physics & motion | 4.47 |
| Subject consistency | 4.94 |
| Audio & sync | 5.00 |
| Average attempts to a usable clip | 1.00 |
| Average generation time | 97s |
| Average cost per run | $1.360 |
| Scored runs | 40 |
By prompt category
| Category | Score | Runs |
|---|---|---|
| On-screen text | 5.00 | 4 |
| Multi-shot | 5.00 | 4 |
| Lip-sync | 5.00 | 4 |
| Image to video | 5.00 | 4 |
| Style | 4.92 | 4 |
| Faces | 4.88 | 4 |
| Animals | 4.79 | 4 |
| Camera | 4.71 | 4 |
| Liquids | 4.67 | 4 |
| Hands | 4.38 | 4 |
Cite this
Free to cite and reuse with a link back. Data is updated as new runs and prices come in, so include the date.
SlopTV. (2026). FLUX 3 benchmark results. Retrieved October 4, 2026, from https://sloptv.co/model/flux-3<a href="https://sloptv.co/model/flux-3">FLUX 3 benchmark results</a> (SlopTV)The clips
Every score above links to the actual generation it came from. Nothing here is a mockup.
Hands avg 4.38/5
Chopping a pepper
Realistic two-handed grip and the knife does pass through the pepper, but slices multiply and reshape between cuts and a loose ring appears on the board around 00:05. Verified frame by frame.
Counting fingers
Five digits throughout, four fingers plus the thumb, verified frame by frame. The thumb opens first and the other fingers unfold roughly in order rather than strictly one at a time.
Fretting a guitar chord
Fingertips press the strings cleanly and the hand shifts between chord shapes with stable geometry, with fitting acoustic guitar audio.
Tying shoelaces
Plausible two-handed lacing with correct finger counts, but there is no finished bow by the end of the 8 seconds and the lace motion goes soft during the crossings.
Faces avg 4.88/5
Face in a crowd
The man stays still and sharp while a dense crowd passes coherently behind him, but the background is not out of focus as asked: the passers-by are nearly as sharp as the subject. Verified frame by frame.
Elderly portrait, slow blink
Deep wrinkles stay sharply defined under the side light and one complete, slow blink closes and reopens the eyes.
Glasses reflection
Round wire glasses with window reflections that shift across the lenses as the head turns. Checked frame by frame; minor softness at the reflection edges.
Turn and smile
Turns from profile to camera with identity fully intact and a natural smile that reaches the eyes.
Liquids avg 4.67/5
Pouring coffee
Continuous slow-motion pour, the level rises plausibly and backlit steam rises throughout.
Honey drizzle
Thick, slow ribbons that pool and fold over the pancakes like real honey; slight morphing at one drip edge late in the clip.
Ice melting in whisky
Photorealistic glass and whisky, but little real melting: comparing the first and last frames, the cube does not clearly shrink, only its top edge softens slightly while condensation thickens. Verified frame by frame.
Wave on rocks
Chaotic, non-repeating spray and foam against stable dark rocks under an overcast sky.
On-screen text avg 5/5
Book chapter heading
"CHAPTER ONE" stays perfectly legible and stable as the camera pushes in. Verified frame by frame.
Neon sign OPEN 24H
"OPEN 24H" is spelled correctly in every frame, flickering in rain as asked. Verified frame by frame.
Bakery sign
"FRESH BREAD DAILY" reads correctly in a painted style throughout the push-in. Verified frame by frame.
Printed t-shirt
"SLOP" stays legible on the moving t-shirt the whole way, the hardest text case in the battery. Verified frame by frame.
Multi-shot avg 5/5
Cafe, three shots
Three shots in the asked order, wide, medium, then hands on the cup, with the same woman in the red coat and the same cafe throughout. Checked frame by frame.
Same bench, day and dusk
Midday then dusk on the same bench with matching framing, ironwork and background trees.
Watch, wide then macro
Wide then macro on the identical watch: same dial, markers, date window and bracelet under the same studio light.
Alley, two angles
From behind down the alley, then front-on as he turns the corner, same man and same mustard jacket. Checked frame by frame.
Lip-sync avg 5/5
Bearded man, long line
Audio says the exact line, transcribed word for word, and the bearded jaw articulates it without warping.
Direct address to camera
Audio says the exact line, transcribed word for word, with eye contact and lips closing on the p of slop.
Sung line
Sung "hold on to me", confirmed in the audio, with sustained open vowels and visible breath.
Two-person exchange
Both lines present in the audio in order; the listener's mouth stays still while the other person speaks.
Camera avg 4.71/5
Crane to skyline
Steady vertical rise from street level that reveals a stable golden-hour skyline.
Dolly in to a door
Slow, steady push-in; perspective opens correctly and the door stays centred, closed and in proportion.
Handheld follow
Natural handheld shake following a child through tall grass, with coherent running motion.
360 orbit
FAILURE MODE: the camera travels only about 90 to 120 degrees, from the front to her side, and never completes the 360-degree orbit. The room itself holds together. Verified frame by frame.
Animals avg 4.79/5
Pigeon takeoff
The pigeon crouches and lifts off cleanly, feet separating from the stone, with wings fully extended at the top of the stroke.
Cat jumping to a counter
One continuous leap from the floor onto the counter with a natural landing; the take-off frames smear with motion blur. Checked frame by frame.
Dog shaking off water
Wet, matted fur and droplets flying outward in a plausible spray; the shake is strongest in the head and upper body rather than the whole body.
Galloping horse
Clean side profile with four distinct legs and hooves meeting the ground, though the gait reads more like a canter than a full gallop. Checked frame by frame.
Style avg 4.92/5
Anime walk cycle
Clean anime style and a face and outfit that never change, but the camera tracks her so she walks in place while the background scrolls, rather than crossing the frame. Verified frame by frame.
Claymation wave
Reads as real stop-motion clay with visible fingerprints and tool marks while the figure waves.
1970s film look
Convincing 1970s look, with grain, halation and warm period colour on a vintage coupe along a coastal road.
Rotating mug
Plain white mug keeps its form while the handle travels smoothly around it, never jumping or duplicating.
Image to video avg 5/5
Animate a landscape
Clouds drift and grass moves while the hills stay put, matching the source photo. Compared with the reference image.
Animate a vintage photo
Stays black and white with the original grain and tonality, adding only subtle motion and a slight push-in. Compared with the reference image.
Animate a portrait
Same face as the source photo from first to last frame, with a slow natural blink and slight hair motion. Compared with the reference image.
Product photo to spin
About 180 degrees of rotation, the handle moving from right to left, with plausible, rigid geometry for the unseen side. Compared with the reference image.