Black Forest Labs
FLUX Kontext is the most precise AI photo editor available — it understands exactly which elements to change and which to preserve, making it the best choice for targeted, surgical edits. Where GPT Image 1 excels at abstract intent ("make this look more professional"), FLUX Kontext excels at literal, specific transformations ("change the red shirt to a blue blazer," "swap the background to a beach at sunset"). Two quality tiers are available: Pro (6 credits/image) for standard edits and Max (12 credits/image) for edits requiring fine detail preservation. Both support 1-4 output images and 9 aspect ratios. For edits combining multiple source photos, the FLUX Kontext Multi variant accepts multiple reference images simultaneously.
Add an Image node containing your source photo, then add a second Image node with FLUX Kontext selected. Connect the first node's output to the second node's input — this tells FLUX Kontext to use your photo as the editing base. The key concept: FLUX Kontext is "edit-mode preferred," meaning when it receives an input image it automatically operates as an editor rather than a generator. You don't need to toggle any mode switch — just connecting an input image activates the editing behavior.
The most common mistake with FLUX Kontext is writing a full scene description instead of an edit instruction. Wrong: "A person in a blue blazer standing in front of a modern office." Right: "Change the red shirt to a blue blazer. Keep everything else exactly as is." The difference matters because a full description tells the model to generate a new image, while an edit instruction tells it to modify the existing one. Start your prompt with an action verb: "Change," "Replace," "Remove," "Add," "Transform," or "Swap." Then explicitly state what should remain unchanged: "Keep the person's face and hair exactly as they are" prevents the model from making unwanted adjustments to protected elements.
Pro (6 credits) handles background swaps, color changes, object addition/removal, and style transfers with excellent quality. Max (12 credits) is worth the 2x cost in two specific cases: edits involving faces (where Pro may slightly alter facial features) and edits to fine textures (fabric weave, wood grain, metal brushing). For a batch of 10 photo edits, the cost-effective approach is Pro for all initial edits, then re-run any face-critical or texture-critical results in Max. A background swap — the single most common photo edit — is virtually identical between Pro and Max.
Complex transformations work best as sequential steps rather than one sprawling prompt. Chain multiple FLUX Kontext nodes: Node 1 swaps the background → Node 2 adjusts the lighting to match → Node 3 applies a color grade. Each node receives the previous node's output as its input, maintaining visual continuity. This chaining approach is more reliable than a single prompt like "Change the background, adjust lighting, and apply vintage color grade" because each step can be individually verified and re-run if needed. It's also how professional retouchers work in Photoshop — layered, non-destructive edits rather than one-shot transformations.
Background swap — the most common photo edit. The explicit preservation instruction ("Keep the person and their outfit exactly as they are") is critical: without it, FLUX Kontext may slightly adjust the subject's appearance to better "fit" the new environment. This prompt works for any background swap — just replace the environment description.
Change the background to a modern minimalist office with floor-to-ceiling windows and city skyline view. Keep the person and their outfit exactly as they are.
Time-of-day transformation — FLUX Kontext handles relighting exceptionally well because it understands how light sources affect an entire scene. "Warm street lights" and "neon signs reflecting on wet pavement" give the model specific light sources to work with, resulting in physically plausible night lighting rather than just darkening the image.
Transform this daytime photo into a nighttime scene. Add warm street lights, neon signs reflecting on wet pavement, and a moody urban atmosphere. Preserve all subjects.
Style transfer — this demonstrates FLUX Kontext's ability to apply aesthetic transformations without changing content. The specific style markers ("warm color cast, film grain, faded highlights, soft vignette") prevent the model from making its own artistic choices. Compare with GPT Image 1: if you wrote "make this look vintage," GPT would infer these details automatically. FLUX Kontext works better with explicit specifications.
Apply a vintage 1970s film photography style — warm color cast, film grain, slightly faded highlights, soft vignette around the edges. Keep the original composition.
Pro (6 credits) vs Max (12 credits): use Pro for background swaps, color changes, and object edits. Use Max only for face-critical edits and fine texture modifications. For most photo editing workflows, Pro delivers identical results at half the cost.
Always start prompts with action verbs ("Change," "Replace," "Remove," "Add") and end with preservation instructions ("Keep everything else exactly as is"). This two-part structure — what to change + what to protect — produces the most reliable results.
FLUX Kontext Multi (same Pro/Max pricing) accepts multiple reference images. Use it for composite edits: combine a person from photo A, a background from photo B, and a product from photo C into a single scene.
For the cheapest photo editing, Qwen Image Edit costs only 4 credits per edit — but with less precision on complex transformations. Use Qwen for simple batch edits (color corrections, basic object removal) and FLUX Kontext for precision work (face-aware edits, brand-critical modifications).
FLUX Kontext is the precision photo editor — it preserves subject identity at approximately 95% fidelity while transforming exactly the elements you specify. This makes it the best choice for targeted edits where you know exactly what you want changed. The three Martini photo editors each serve a different editing style: FLUX Kontext for precise, literal edits ("change X to Y"); GPT Image 1 for abstract, intent-based edits ("make this look more professional"); and Qwen Image Edit for high-volume, budget-friendly simple edits at 4 credits each. For professional retouching workflows, FLUX Kontext's node-chaining capability on the Martini canvas effectively replicates Photoshop's layer-based editing — but in seconds rather than hours.
Connect FLUX Kontext with other AI models on Martini's infinite canvas. No GPU required — start free.
Get Started FreeOpenAI
GPT Image 1 is the most intelligent image editor on Martini — it understands abstract, intent-based instructions that other models cannot interpret. Where FLUX Kontext and Qwen require precise, literal edit commands ("remove the red car"), GPT Image 1 understands conceptual requests like "make this look more professional," "add a festive holiday feel," or "make this photo feel warmer and more inviting." This language understanding comes from GPT's foundation as a language model, giving it an editing capability that feels closer to working with a human designer than an AI tool.
View guideAlibaba
Qwen Image Edit is the cheapest and fastest image editor on Martini — just 4 credits per image with near-instant results. It excels at object-level edits: adding, removing, replacing, or recoloring specific elements with clean, natural blending. Where GPT Image 1 understands abstract concepts ("make it more professional"), Qwen works best with concrete, literal instructions ("replace the red car with a blue bicycle"). This specificity paired with low cost makes it the go-to choice for batch editing workflows where you need to process 10-50 images quickly.
View guide