Prompt Engineering

Budget: 999 UAH Deadline: 2 days

Hello! I generate images through Gemini every day, so I'll say right away the main point: the style does not consistently hold with just the text; the model interprets the words a bit differently each time.

Another scheme works — alongside the prompt, there is a reference image, and the prompt itself is divided into two parts: an unchanging style block and a changing subject block, which you adjust for each generation. The similarity of a person in the photo is done the same way with a reference, not a description of the face; otherwise, it results in a "similar relative."

I will gather such a template for you and show it with a few examples to demonstrate that the style does not drift. I will take your budget, two days. Is the style already present in the finished images, or does it also need to be derived?

Budget: 1000 UAH Deadline: 1 day

Hello!

I can help with this task, but it's important to break it down into several components.

What you are describing cannot be solved with one universal prompt. Usually, it involves a separate system of prompts:

* Style prompt — describes the artistic style, color palette, materials, lighting, level of detail, and other unchanging parameters.
* Character Card — a separate instruction that captures appearance, proportions, clothing, facial features, and other permanent characteristics. This helps maintain maximum consistency of the character across different scenes.
* Prompts for specific scenes — created based on the style and character card. A master scene can also be used, which sets the overall atmosphere and serves as a base for further generation.

I can help with:

* developing a system of prompts for Gemini (or suggest another model if it better suits your task);
* creating a character card template;
* writing instructions on how to work with the prompt system to ensure the results are as predictable and stable as possible.

To assess the task, I would like to get a bit more information about the project:

* what specific style needs to be maintained (are there references?);
* which model do you plan to use (Gemini, Imagen, Midjourney, FLUX, etc.);
* how many characters and scenes are planned;
* whether character generation from photographs of real people is needed.

After that, I will be able to suggest an optimal structure for the prompts and evaluate the scope of work. Expand

Budget: 999 UAH Deadline: 3 days

One text prompt alone will not capture the style and identity of the character. A breakdown of the reference, structure of parameters, negative constraints, and test scenarios are needed. In 3 days, I will prepare a master prompt for Gemini, variant blocks for objects, people, and scenes, negative constraints, 5-7 test cases, and instructions for references. I cannot promise absolute similarity to a person with just a prompt, but repeatability can be significantly increased through a reference frame and a fixed scheme. Andriy, do you have 3-5 examples of the desired style and which versions of Gemini are available to you?

Budget: 999 UAH Deadline: 3 days

Hello, Andriy!

The task is clear: you need to build a projected pipeline and prompt structure to maintain a consistent visual style, preserve the facial features of characters from photos (Face Consistency / Character Sheet), and ensure quality integration of objects into the scene.

How I can help and the approach I propose:
Working with Gemini / Imagen 3 and alternatives:

Gemini (Imagen 3): I will create a systematic prompt with a clear specification of the stylistics (lighting, textures, palette, color space, rendering), as well as a methodology for referencing.

Alternative solutions (if greater control is needed): Midjourney (Style Reference --sref, Character Reference --cref) or Stable Diffusion (ControlNet + InstantID / IP-Adapter for 100% accuracy in preserving the character from the photo).

What you will receive as a result:

Template/Prompt Builder: A ready structure (Master Prompt) into which you can simply insert the necessary objects, actions, or people without losing style.

Guide/Instructions: A document with working rules, key parameters, and recommendations on how to avoid "artifacts" and achieve predictable results.

Test Generation: Conducting tests on your references and character photos for final approval. Expand

Budget: 999 UAH Deadline: 3 days

I will do it. 11111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111

Budget: 1002 UAH Deadline: 2 days

Hello. I will create a system prompt for you that captures the style and ensures character similarity with the references. I work with Gemini — I will select the optimal structure for the expected result without "hallucinations." I will develop an algorithm for using Image Reference so that new objects fit organically into the given scene. You will receive a ready-made master prompt and a short instruction on how to work with it for scaling. I will do one test generation based on your example to confirm accuracy. Message me privately to discuss the project.

Budget: 999 UAH Deadline: 2 days

I will create a prompt system, not just one prompt, because you have two different tasks: to maintain style and to preserve the likeness of a specific person. One text reliably achieves only the first.

I fix the style with a separate prompt and your example reference so that it does not drift between frames. The likeness of a person from a photo is no longer described in words, but is a separate photo-face that the model holds in different scenes.

Here, the model is key, and I do not blindly pull one. I will test the two strongest on your references: Gemini Nano Banana Pro and GPT Image 2.0. Each excels in different areas, so I will create several frames in both, show where each better maintains style and likeness, and provide a reasoned recommendation, not just an opinion in words. Both deliver this immediately, without separate programs and unnecessary settings.

I will provide ready-made blocks of prompts (style, character, scene, negative constraints), instructions on how to substitute new photos, and a small test set of frames so you can immediately see when the series drifts.

How many style examples do you have? I will test them on those.

Budget: 999 UAH Deadline: 5 days

I have seen similar tasks related to generation where a stable, predictable result is needed, so I understand what is being discussed.

Specifically, the task here is to fix the stylistics while being able to substitute different objects or characters (especially based on photos) without losing the integrity of the scene. This is achieved through a well-constructed system prompt with a detailed description of the style, as well as techniques like reference image weighting, negative prompting, and seed control. For characters based on photos, Gemini Imagen 3 or Ideogram with img2img are suitable, but I would also consider Stable Diffusion with LoRA or IP-Adapter if a really high similarity to a real person and control over the result is needed.

I can test several approaches, gather comparative results, and create a final prompt with usage instructions. I am ready for details, please write if there are examples of style or references.

Budget: 999 UAH Deadline: 3 days

Good day!

I work with Gemini, MidJourney, and Leonardo, so I would like to try to solve your task. I understand that it's not just about writing a prompt here, but testing different options to ensure the style is preserved as much as possible when generating new scenes and characters. If you already have references, I can start working right away.

Budget: 999 UAH Deadline: 3 days

Good day!

I understood the task — it's not just about simple generation, but a stable approach: we set the style once, and it doesn't "wander" from image to image, even when we substitute new characters or objects. I'm also interested in the aspect of photos of people — so that the stylized character is still recognizable and fits well into the scene, rather than looking like it's just stuck on.

I would do this on Nano Banana 2 (the current Gemini model for images) — it maintains character similarity well when changing style and environment, without external plugins. I will create a prompt template, test it on your references, and provide a ready instruction on how to substitute new photos in the future.

Budget: 1000 UAH Deadline: 2 days

Hello.
Write everything, I will do it.
It is necessary to create a database for skills for AI.

Budget: 1000 UAH Deadline: 1 day

Good day, I will gladly help you. I look forward to collaborating. I also have experience with construction and drafting.

Budget: 2500 UAH Deadline: 1 day

Hello! Achieving stable results with just random text is difficult, but the task can be solved with a well-structured prompt and working with references in Gemini.

For 2500 UAH, I will develop a ready-made prompt template tailored to the required style, test it on your examples of people and objects, and provide a brief instruction on how to use it correctly. As a result, you will be able to simply substitute new photos and consistently receive predictable generation in a unified style.

I am ready to discuss the task in more detail, feel free to write.

Budget: 999 UAH Deadline: 1 day

An interesting task, exactly what I'm working on.

The style is fixed through a master prompt plus a reference frame and negative constraints — then the result is predictable even in series. Gemini is possible, but for facial similarity, Flux or Midjourney are more stable.

I will prepare a prompt system tailored to your style, templates for objects and characters, and a short guide to ensure the series remains consistent.

What references do you already have?

Budget: 990 UAH Deadline: 1 day

Hello. I will take on your task, set up image generation in one style, and develop a prompt to maintain facial similarity with the photos. Do you already have specific style references to which the characters need to be adapted? Message me privately, and we will discuss all the details and determine the deadlines. I will be happy to help)

Budget: 1000 UAH Deadline: 1 day

Good day.
I am ready to develop a prompt system for generating images in a fixed style with predictable results.
Portfolio: https://gooru-lab.github.io/portfolio/
Experience: prompt engineering for LLM/AI, structured instructions "do not deviate from the rules," iterative quality testing of output.
What I will do:
• I will capture your style (references → style bible / master prompt);
• prompt templates: objects, people/characters from photos (similarity + integration into the scene without changing the style);
• rules for negative prompts and a checklist to verify "style has not drifted";
• model recommendation: Gemini (if suitable for the task) or an alternative (e.g., Flux / Midjourney / Ideogram) — with justification;
• a short guide: how to generate a series of frames consistently.
To start: 5–10 style references + 1–2 photos of people/example scene. Deadline: 3–5 days after references. Expand

Budget: 2000 UAH Deadline: 4 days

For a truly fixed commercial style:
Stack
ComfyUI
FLUX or Stable Diffusion
Style LoRA — trained on your style
Character LoRA — separately on a specific person
IP-Adapter / FaceID — retains the face
ControlNet Pose — fixes the pose
ControlNet Depth/Canny — fixes the composition and shapes
ControlNet allows you to separately fix the contours, depth, and structure of the scene. For Stable Diffusion 3.5, Canny, Depth, and Blur ControlNet are officially available.

THIS is not one model and one Prompt, THIS is a combine, it can be applied in different ways.
You will need to set up the pipeline and teach you. Expand

Budget: 999 UAH Deadline: 2 days

Ready to create a universal prompt system for you with a clear structure for style fixation and maintaining similarity of faces/characters.

I will develop a ready prompt template and a step-by-step guide on how to adapt it for Gemini and alternative tools (Midjourney / Flux) to achieve a predictable result.

Please let me know if you have examples (reference images) of the same style that needs to be fixed, and what level of similarity to a real person you require—100% photographic or stylized recognition?

Budget: 999 UAH Deadline: 2 days

Good day! I have completed similar projects for myself and clients, I understand how to adapt prompts for different neural networks and control the generations. In case of collaboration, I would like to receive a detailed technical assignment and references. Feel free to write, I am always in touch, I don't inflate prices, and I meet deadlines. Portfolio: Freelancehunt

Similar projects: https://freelancehunt.com/showcase/work/stvorennya-vizualizatsiyi-dlya-knigi-retseptiv/2024057.html https://freelancehunt.com/showcase/work/realistichniy-mokap-pufa-kimnati/2033478.html

Budget: 999 UAH Deadline: 3 days

Andrew, you want to achieve a stable visual result so that the AI not only draws similar objects but clearly maintains the specified style and anatomy when integrating characters. For this solution, we will focus on creating a systematic prompt with fixed style parameters and setting control points for generation. This will make the process of creating new scenes predictable rather than random. Which specific stylistic direction is a priority - photorealism, illustration, or something else where the preservation of integrity is most important?