Imagen 4 vs Flux 1.1 Ultra: Photorealism Showdown
We ran identical prompts through Imagen 4 and Flux 1.1 Ultra on the 8frame canvas. Here's the honest breakdown, per-use-case picks, and the pricing math.
Imagen 4 wins on prompt adherence and lighting physics. Flux 1.1 Ultra wins on edge sharpness and generation speed. For most photorealistic production briefs, that difference is real and it will show up in your output, especially when the prompt specifies directional light or fine surface detail. Here is what we observed running both on the 8frame canvas with identical prompts.
TL;DR
- Best light physics and prompt adherence: Imagen 4 at ~$0.05 per image, ~10 second generation
- Best edge detail and sharpness at 4K: Flux 1.1 Ultra at ~$0.04 per image, ~5-10 seconds
- Speed: Flux 1.1 Ultra is consistently faster, 5 seconds vs 10 for Imagen 4
- Verdict: Imagen 4 for architecture and portrait work where directional light matters; Flux 1.1 Ultra for product and landscape where edge resolution is the priority
The test setup
Both models were run on the 8frame canvas in June 2026, default settings, no post-processing, no custom samplers. We used four prompts covering the main photorealism use cases: architecture, portrait, product, and landscape. Same prompt text, same resolution target (4K), different model nodes side by side.
The comparison here is Imagen 4 vs Flux 1.1 Ultra specifically. For a broader comparison that includes Nano Banana Pro, Seedream 5.0, and the Flux budget tiers, see nano banana vs seedream vs flux.
Photorealism test prompts and results
Architecture
A minimal concrete house exterior at blue hour, large fixed windows with warm interior light visible from outside, a single path light casting a cone on the stone walkway, surrounding pine forest soft-focused in background, architectural photography, shot on Hasselblad X2D, 28mm f/5.6, ISO 200, 4K
Imagen 4: The interior warm-to-exterior cool transition read correctly across the window pane. The path light cone had a believable falloff on the stone surface with a visible penumbra. Forest background defocused at a depth consistent with the specified focal length. The strongest result across both models on this prompt.
Flux 1.1 Ultra: Edge detail on the concrete surface was sharper, and stone texture on the walkway was more granular. But the path light cone fell straight rather than showing the slight diffusion a real fixture produces, and the interior-to-exterior color temperature transition was slightly compressed. Still production-usable, but the light physics were looser.
Portrait
A woman in her early 40s, indirect overcast window light from the right, no fill, high cheekbone shadow, strand-level hair detail in a loose bun, subtle skin texture, no retouching aesthetic, documentary portrait, shot on Sony A7RV, 85mm f/1.4, ISO 200, 4K
Imagen 4: Shadow placement matched the specified overcast right-side source, the shadow under the cheekbone read as diffuse rather than hard. Skin pore detail was present without smearing. Hair strands were individually resolved in the highlight areas. The "no retouching aesthetic" instruction held.
Flux 1.1 Ultra: Hair strand resolution was tighter and more detailed than Imagen 4, which is Flux's consistent strength on texture-dense surfaces. The shadow, though, was slightly more dramatic than an overcast source justifies. The cheekbone shadow had more contrast than a soft window would produce. For studio-controlled portraits where the light source is specified, this is a meaningful deviation.
Product
A glass perfume bottle on a white acrylic surface, single overhead softbox at 45 degrees, caustic light refraction through the glass visible on the surface below, white background, product photography, shot on Phase One IQ4, 120mm macro f/8, ISO 50, 4K
Imagen 4: Caustic refraction appeared on the surface, though it was slightly idealized, more regular and symmetrical than a real glass refraction pattern. The bottle glass read as glass, the overhead shadow was correctly positioned at 45 degrees.
Flux 1.1 Ultra: Edge resolution on the bottle was noticeably sharper. The silhouette against the white background held without the slight edge softening Imagen 4 showed. Caustic refraction was less accurate than Imagen 4 but surface sharpness was higher. For product work where the brief is clean lines against white, Flux 1.1 Ultra's edge advantage is worth having.
Landscape
A volcanic black sand beach at sunrise, first light catching the wave crests, dark lava rock formations in the mid-ground with foam pooling in the crevices, no people, wide landscape, shot on Canon R5, 24mm f/8, ISO 100, 4K
Imagen 4: Wave crest highlights held detail without blowing out, foam in the rock crevices was irregular and physically plausible. The dark lava texture differentiated from the sand.
Flux 1.1 Ultra: Sharper grain on the sand texture, stronger micro-contrast on the lava rocks. Wave motion was less physically convincing than Imagen 4's, the crest highlights looked slightly painted. But the still-texture quality was higher. For landscape work that ends up at billboard scale or print, Flux 1.1 Ultra's resolution advantage matters more.
Imagen 4 strengths and weaknesses
Strengths. Prompt adherence is the clearest strength. When you specify a light source position, Imagen 4 follows it. When you specify a surface material, the material reads correctly. Shadow physics on directional sources (window light, path lights, softbox at angle) are the most accurate of the two models. Portrait lighting is reliable: soft source stays soft, hard source stays hard.
Weaknesses. Slower, typically around 10 seconds per generation. Edge sharpness at 4K is slightly below Flux 1.1 Ultra on high-contrast surfaces. On the product photography test, caustic glass refraction was accurate but idealized, slightly more regular than a real glass object produces. Also priced at ~$0.05 per image, one cent higher than Flux 1.1 Ultra.
Flux 1.1 Ultra strengths and weaknesses
Strengths. Edge sharpness is the clearest strength. Hair, fabric weave, stone texture, glass silhouettes against white, all resolve at a higher resolution than Imagen 4. Generation speed is faster, consistently 5-8 seconds in the 8frame queue. Slightly cheaper at ~$0.04 per image. For volume production where you're running batches of 20-50 images, the speed and price gap compounds.
Weaknesses. Light physics are looser on directional sources. On portrait prompts, shadow contrast runs slightly hotter than the specified source justifies. On architectural prompts, interior-to-exterior color temperature transitions are compressed. For anyone who has worked with a lighting director on a real shoot and can spot a misread shadow, this will land differently than it does on a stock image.
For prompt strategies that get the most from Flux 1.1 Ultra, see flux 1.1 ultra prompts for photorealism.
Best by use case
Architecture visualization
Imagen 4. Light physics matter more here than edge sharpness, and Imagen 4's handling of mixed interior and exterior light is the more accurate of the two. Archviz clients who have looked at real buildings can tell when the path light falloff is wrong. That distinction costs you client confidence.
Portrait photography
Imagen 4. Soft sources stay soft, hard sources stay hard. The "no retouching aesthetic" instruction holds more consistently. For documentary or character work where the lighting setup is part of the brief, Imagen 4's shadow accuracy is the deciding factor.
Product photography
Flux 1.1 Ultra for clean-line products; Imagen 4 for glass and transparent objects. On hard-edged products against white or neutral backgrounds, Flux 1.1 Ultra's edge resolution wins. On glass, crystal, or any object where refraction accuracy is the brief, Imagen 4 is stronger. If your product line is both, run one generation of each and pick per product.
Landscape and environment plates
Flux 1.1 Ultra. For work that ends up large-format, print, or billboard scale, the micro-contrast and grain detail matters more than the motion physics in a still. Flux 1.1 Ultra's texture resolution holds better at 100% zoom on wide-format outputs.
Pricing math
These are June 2026 numbers from the 8frame canvas.
| Model | Cost per image | Resolution | Generation time |
|---|---|---|---|
| Imagen 4 | ~$0.05 | Up to 4K | ~10s |
| Flux 1.1 Ultra | ~$0.04 | 4K | ~5-10s |
At 100 images per month, Imagen 4 costs $5 more. At 1,000 images per month, that is $10 difference. Neither number is a reason to compromise on the model that fits the brief. Where the difference becomes meaningful is in automated pipelines where you are generating 5,000 or more images per month, and in those cases, Flux 1.1 Ultra's speed advantage (faster generation means lower queue time and faster pipeline throughput) matters as much as the cent-per-image difference.
If you are deciding between these two specifically for a volume use case, Flux 1.1 Ultra is the sensible default for most product and landscape generation pipelines. For portrait and architecture pipelines where a human reviews each output before delivery, Imagen 4's accuracy is worth the extra cent.
Step-by-step: running both models on 8frame
- Open the 8frame canvas and add two image generation nodes side by side.
- Assign one node to Imagen 4, the other to Flux 1.1 Ultra.
- Connect both nodes to the same prompt input so the prompt propagates identically.
- Set both to 4K output.
- Run a single generation and compare outputs before iterating.
- For the model that gets closer to your brief, lock that node and run 3 variants to pick the strongest.
The side-by-side structure takes about two minutes to set up and saves the iteration time you would otherwise spend re-running one model hoping for a different result.
FAQ
Is Imagen 4 better than Flux 1.1 Ultra for photorealism?
It depends on what kind of photorealism. For directional lighting accuracy and prompt adherence, especially on architecture and portrait work, Imagen 4 is stronger. For edge sharpness, texture resolution, and speed on product and landscape work, Flux 1.1 Ultra is competitive or better. The two models have different strengths in the same photorealism category.
How much does Imagen 4 cost per image?
Approximately $0.05 per image at 4K on the 8frame canvas. Flux 1.1 Ultra is approximately $0.04 per image. Both are available from a single 8frame account. For a full model pricing comparison that includes Nano Banana Pro, Seedream, and the Flux budget tiers, see nano banana vs seedream vs flux.
Which model is faster, Imagen 4 or Flux 1.1 Ultra?
Flux 1.1 Ultra is faster. It consistently generates at 4K in 5-8 seconds on the 8frame queue. Imagen 4 runs around 10 seconds. For single-image creative work the difference is minor. For automated pipelines running hundreds of images, the throughput gap is meaningful.
Run the Imagen 4 vs Flux 1.1 Ultra comparison yourself from a single canvas node. Same prompt, both models, outputs side by side.