FireRed Image Edit 1.1 was tested as a practical one-image editor, not as a beauty-shot generator. The question was simple: can it change a person, a product, or a landscape while preserving the part that makes the image usable? Six existing runs cover identity, accessories, product edges, materials, seasonal transformation, and night relighting.
FireRed Image Edit 1.1 on Wiro accepts an input image plus a written edit instruction. The Wiro model page exposes steps, guidance scale, sample count, seed, width, and height. This set used 30 steps, guidance scale 1.5, one sample, seed 0, and 1024 x 1024 output settings. The source files are not square in every case, so the visible output framing remains part of the result rather than a promise of exact source dimensions.
What this FireRed Image Edit 1.1 test checks
A useful image edit has two jobs: make the requested change and leave the unrelated evidence alone. Portrait edits need a recognizable face. Product edits need clean boundaries, readable design, and plausible contact shadows. Landscape edits need stable silhouettes and believable light. The six prompts deliberately move from local changes to broad scene changes so the weak spots are harder to hide.
FireRed’s makers position version 1.1 as an update for portrait consistency, multi-image conditioning, and specialized editing. The published model card also describes text-style reference, restoration, and multi-element work. This page only claims what these six one-image outputs show. For the broader release material, see the official Hugging Face model card and the FireRedTeam GitHub repository.
Settings, time, and cost
| Input mode | One source image plus one text instruction |
| Steps | 30 |
| Guidance scale | 1.5 |
| Samples | 1 |
| Seed | 0 |
| Requested output size | 1024 x 1024 |
| Documented Wiro example runtime | 6.0000 seconds for one output |
| Documented Wiro example cost | $0.003510000000 total in the docs cancellation example |
The last two figures come from the Wiro model documentation, not a fresh benchmark. The documentation shows a six-second completed example and separately shows the listed total cost in a cancellation payload. Queue time, input size, output size, and deployment conditions can change both figures, so this is a traceable example rather than a fixed price quote.
Six before-and-after outputs
1. Silver bob, studio light, and a plain background
Prompt: Change hair to short silver bob haircut. Add soft studio rim light. Keep face identity. Remove background and use solid light gray.
 |
 |
This is a compact identity test. The output combines a hairstyle replacement, relighting, and background removal without asking for a new pose. It is the right kind of task for profile variants, but a production portrait still needs a face and hairline review at full size.
2. Glasses, a red jacket, and night bokeh
Prompt: Add thin round eyeglasses. Change jacket to bright red bomber jacket. Keep face identity. Background: blurred city lights at night.
|
|
 |
The second output puts three edits in different image regions: eyewear near the face, clothing across the torso, and an entirely new background. That makes it more demanding than a simple color swap. It is useful for editorial mockups and campaign concepts, but not for a final product claim where garment construction must be exact.
3. Watch cutout with a white seamless
Prompt: Replace background with clean white seamless. Add soft shadow under the watch. Keep product shape and logo.
 |
 |
This output tests whether the object stays grounded after the background disappears. The soft shadow makes the watch read as a product photo rather than a floating cutout. It is a sensible use for early catalog layouts, while logos, dial marks, and edge masks should be checked before ecommerce export.
4. Brushed silver material on slate
Prompt: Change watch body to brushed silver metal. Keep dial layout. Background: dark matte slate.
|
|
 |
Material replacement is less forgiving than a background swap because reflections can overwrite tiny details. The output is useful for concept direction and art boards. Choose a manual retouching pass instead when every bezel mark, engraved word, or product finish must match the SKU.
5. Winter conversion
Prompt: Turn scene into winter. Add fresh snow on the rocks. Make light cold and overcast. Keep composition.
 |
 |
This is a broad-area edit. The prompt asks the model to change rock texture, ground coverage, and the sky mood while retaining composition. That makes it a strong fit for concept frames and travel-story treatments, where overall atmosphere matters more than forensic consistency.
6. Aurora night relight
Prompt: Convert to cinematic night scene with aurora. Add stars. Keep mountain silhouette. Keep composition.
|
|
 |
The final output is the most aggressive scene rewrite. It tests sky replacement and color spill along a hard silhouette. It works best as a fast creative direction tool; inspect mountain edges and any foreground detail before using it in a composited final.
When to pick FireRed Image Edit 1.1
Pick FireRed Image Edit 1.1 when one input needs several connected edits at once: a portrait plus wardrobe and setting, a product plus surface and background, or a landscape plus weather and time of day. The documented controls make it practical to keep a repeatable starting point: 30 steps, scale 1.5, one sample, seed 0, and 1024 square output.
Pick a narrower workflow when the deliverable has zero tolerance for drift. Product logos, legal packaging text, exact garment construction, and regulated before-and-after evidence need human review or traditional retouching. For a direct comparison with another editing model, read FireRed Image Edit 1.1 vs FLUX.2-dev. For a different one-image editing benchmark, see Seedream V5 Lite image editing tests. A larger product-focused set is also available in this image edit API comparison.
Run the six prompts on source images that resemble the real job. That is the fastest way to see whether identity, product detail, or scene structure is the constraint that matters most.
