Concept Art and Character Design
05/05/25 - 24/05/25
Introduction
This blog post explores the concept development for the character I plan to model rig and animate as part of this module. The creative element of this character design and context is explored in my blog post Preproduction Exploration to Prepare for Career Development in the Graduate Portfolio Module (Lawson-Hall, 2025a). This blog post covers ideas development, ideas development, research, analysis, concept art, storyboards and project management to create clear technical and creative direction for this project. This blog post will explore rough silhouette development, initial line art, line art and colour development of both the character and the character's weapon.
In my previous module's academic assignment (Lawson-Hall, 2025b) I explored how artificial intelligence (AI) is being used in the screen industries (games, animation and VFX). Keen to explore this further and in the interest of producing a strong character design I wanted to experiment with the practical use of AI in this workflow. This felt like the best place to use this as I needed a strong character design and I'm not exploring myself as a concept arts but more of a character animator. I also wanted to use my conclusions from my research that AI is best used when employed alongside human artistry. Therefore, I wanted to feed AI my concept art and use it to interpret, refactor and reimagine my designs. Depending on the results I potentially could use this to produce a final character design or reference sheet that could be used for modelling.
Project Management
The below image shows the production schedule showing that I plan to spend 2 weeks on concept art development. This is actually outside of the module's start date, however, I'm keen to get ahead so that I can get to animating as quick as possible. At this moment in time everything is on track, however, there potentially could be some disruption as my teaching role is entering an intense end of year assessment period, therefore I may need to allocate some extra time to that to ease the stress.
This below image shows the Trello card for this section of the character production pipeline broken down into separate tasks. It has a deadline and aims to focus my production efforts to build the design effectively over time.
Character Concept Art Development
This section details the production of the character concept art, starting with silhouette development. Throughout this section I am referring to the preproduction research, mood boards, character profile and other elements outline in my blog post Preproduction Exploration to Prepare for Career Development in the Graduate Portfolio Module (Lawson-Hall, 2025a).
Silhouettes
The below image shows my silhouette development. The plan here was to develop a strong character outline that I could use as a base to develop the line art.

Overall, this has came out well and you can see a clear visual development over time with the last 2 designs being the strongest. The third design on the bottom row has a good pose and strong visual shapes that create contrast. For example you can see large shoulders and a thinner waist/leg balance. This gives the character a top heavy design that suggests strength and athleticism. The fourth design on the bottom row is a little more balance but still features athletic build elements like defined arms, shoulders and legs. Whilst I like the chunky shoulders on the bottom third design, this could cause challenges when animating and I want to create a character that I can animate athletically as they traverse an environment and engage in combat. The bottom fourth design also features interesting spikey thorns on the forearms that tie into the character's plant-like roots from their character profile. Therefore, the fourth design is the one I'm going to work into and use to develop some initial line art sketches.
Thumbnail Line Art Sketches
This section shows the initial thumbnail line art sketches that aim to build the design and personality in the character design. I wanted to incorporate the space ranger suit elements but also plant-like and nature aspects.
The first design was my first initial sketch that I actually produced before my silhouettes. I wanted to experiment and see if I could visualise what weas in my mind. This wasn't very successful due to the proportions being slightly unrealistic, but there were some key elements I liked such as the chest detail, belt and thorns. The other 3 designs were much more successful because they used the silhouette design as a base. I experimented with a range of ideas and details but preferred the design on the far right which balanced the alien-plant-like base with sci-fi space ranger clothing. The gem on the chest was really engaging - this was like a seed or lifeforce that powered the character or gives them strength. The hooked nose made them slightly more alien and interesting but kept the humanoid base meaning it would be more accessible to model and animate through following traditional conventions.
Overall, I was happy with this thumbnail sketch in achieving my original vision for this character (Lawson-Hall, 2025a). The next steps were to refine this into a final design and build the colour.
AI Character Design Development
In this section I explore the use of AI (ChatGPT) by using my own art as a stimulus. It will detail my prompts, the AI responses and my reflections to show the journey across this process. At this point in time I was not sure how successful this would be and if it would be any help.
Firstly I started off by feeding it my thumbnail line art and giving it a prompt that linked to elements of my research to steer it in the right direction. This is reflective of what I would do when developing a concept, through studying the research. The prompt was : "Turn this line art into a character concept art. Use a cartoon stylised art style similar to Ben 10 or Star Trek Prodigy. The character is a plant like space ranger. Use this line art as a guide."
 |
Copy of my line art that was attached as part of my prompt. |
This was ChatGPT's (2025a) response :
This was pretty good! Its understood that I was aiming for a athletic body physique, plant like elements and a sci-fi inspired space ranger outfit. the overall art style is good and fits the target audience that I was aiming for, linked to my previous research (Lawson-Hall, 2025a). I really didn't like the odd kelp-like parts on the forearms, these seemed a little confused and not aligned with my original vision of thorns poking out of the character's skin. Therefore, I gave it a follow on prompt "Good! The style is perfect. Can you make the things on the forearms thorns instead of leaves?"
ChatGPT's (2025a) response :
This was much closer to what I wanted on the forearms, however, the thorns were a little large. Some elements of the design have also lost details such as the yellow stripes on the hips and the knee pads. This shows that whilst AI tools are powerful they are likely to include errors or fix one element whilst breaking another. It was my plan to use this to help visualise my design but to ultimately draw the final concept art myself. This would take into consideration the findings from my exploration but create a custom final design that builds on my original artwork.
Next I wondered if it could include more gardener inspired elements into the design but keep them sci-fi inspired. This was an original design idea I had and featured on my moodboard (Lawson-Hall, 2025a). I gave it the follow on Prompt : "Can we make the space outfit inspired by a gardener outfit? Like Dungarees but sci fi inspired?"
ChatGPT's (2025a) response :
I wasn't keen on this response at all. Whilst it included dungarees that were a common item of clothing worn by gardeners, they were not sci-fi inspired by the slightest, therefore didn't achieve the vision. I tried to steer it a little further using the follow on Prompt : "Great but what would a space gardener outfit look like on this character?"
ChatGPT's (2025a) response :
This was beginning to really lose the space ranger vibes from my original concept and leaned too heavily into the gardener aesthetic. The face and arms had completely changed now too and it was evident that it was taking it in the wrong direction. I took one of the images it gave me previous (pictured below), attached it to the chat and gave it the follow on prompt : "Smaller thorns on the forearms."
 |
Attached image to prompt. |
ChatGPT's (2025a) response :
This design was now heading back in the right direction! The thorns on the forearm were smaller and closer to what I wanted. I didn't want the thorns to be too big because they could cause intersections when trying to pose the 3D character and I wanted as much flexibility as possible whilst still keeping a unique design. It has also evolved the face a little as well without prompt - this looks okay but the nose area is less 'hooked' than other designs, however, this is something I could potentially fix when I draw the character myself.
I started a new chat using a similar prompt to before to see if this would interpret the design differently. I reattached my concept art (shown below) and gave it the prompt : "Turn this line art into a character concept art. Use a cartoon stylised art style similar to Ben 10 or Star Trek Prodigy. The character is a plant like space ranger. Use this line art as a guide."
 |
My concept art reattached to the new chat. |
ChatGPT's (2025a) response :
It did give me a new design, however, the style is very similar and it's understanding the sci-fi inspired outfit from my concept art well. I did like the detail in the space suit this time, it was keeping an element of medieval sci-fi that I thought worked really well. The gem idea on the chest is coming through strong and is a key element to the design. Its reintroduced similar problems to before - such as the kelp-like elements on the fore arms opposed to the original intention of thorns. To redirect the design and maintain the idea of space gauntlets with thorns poking through, I gave it the follow on prompt : "The arms should have thorns on them. Can you make it so he is wearing gauntlets that fit in with the style of the rest of the armour but there are holes for small thorns to go through."
ChatGPT's (2025a) response :
This was working well and achieved the prompt I set out. The overall design was looking a little too bland and I wanted to try experimenting with how it would handle altering the colour palette. I gave it the prompt : "Can you keep the design the same but try a different colour palette?"
ChatGPT's (2025a) response :
The colour palette was a lot more striking and it was interesting to see how it handled this change. You can also see its introduced more thorns and has refactored the outfit so that it has slowly changed and morphed - actually losing a lot of the detail that made the design interesting in the first place. This is a major draw back of AI and shows that human artistry is required to create that final version that would be production ready. Despite this, its really powerful that I can iterate so quickly through ideas and gain a visual representation of what something could look like. This is empowering and results in a lot less wasted time as I consider my artistic ideas. It is also worth noting that I couldn't have got this close without feeding it my own artwork initially, which highlights the value of keeping humans at the centre of the creative process.
I was liking the design, however, the head on previous designs it created were more interesting. I wanted to see if I could get it to blend elements of the two designs together so I gave it the follow on prompt "Can you take elements of this design into the last one you made? I like the shoulder armour, knee armour and the boots on the design attached to this message. I like the gauntlets with thorn on them last image you created. I prefer the head on them image attached to this message." with attached image :
ChatGPT's (2025a) response :
This was working well because it was starting to balance all the right elements that I wanted from the design together. The head now resembled the original generation and my original concept to feature the humanoid like base with alien like ears and nose. There was simplified elements of the space suit outfit with the gauntlets and thorns I wanted. It also had the 'lifeforce' gem on the chest that really makes this design stand out and align with my game concept. I wanted to explore this gem design further and gave it the follow on prompt "I like the blue gem on the chest, can we work that into the shoulder and knee armour designs?"
ChatGPT's (2025a) response :
This was looking really interesting and aligned closer to my original concept that I fed it. I was liking how the knee pads and shoulder pauldrons incorporated these gems, however, the inclusion of them on the gauntlets felt a little too much. There are also balancing issues with the gold detail lines on the chest not being perfectly symmetrical. At this point I was getting close to what I wanted and I was curious if I could use AI to turn the character around. This would really help with the modelling process because I could see the character from multiple angles, with hopefully the proportions remaining the same. To experiment with this I gave it the follow on prompt : "Can you create a rear view of the last character you created?"
ChatGPT's (2025a) response :
This was pretty good! It kept the proportions well, however, did include some graphical errors like the weird half-kelp and half-thorn-like elements on the forearms. This didn't matter too much because I would be able to correct or build on these design elements themselves. I then wanted to see if I could get a side view so tried the follow on prompt : "How about a side view?"
ChatGPT's (2025a) response :
This was also successful from a proportions perspective, however, the design was beginning to fall apart a little bit and start straying a little too far from what I wanted. However, this showed me that once I got to a point where the design was working well I should be able to use this technique to create a reference sheet with multiple views for the character.
I returned to another design that it had generated and wanted to work on the face a little bit. I gave it the follow on prompt "Could we keep this design the same but make the nose a bit more pointy. It should point upwards rather than curving downwards." with attached image :
ChatGPT's (2025a) response :
This wasn't quite what I meant and I realised the prompt wasn't quite where it needed to be to achieve this. It was also beginning to lose a bit of style I originally wanted.
I wanted to experiment with the space suit outfit development but include reference of a face design that I liked. I wanted to try referencing other games from my research to see if this steered the design any closer to what I wanted. I created the follow on prompt "Could you keep the blue gem in the middle but take inspiration from Ratchet’s sci-fi outfit from the Ratchet and Clank series? I was imagining a little more metal and sci-if suit elements." with attached image :
ChatGPT's (2025a) response :
This design was really strong and the spacesuit had a lot more interest to it, as well as looking practical for combat. I was a little worried about how restrictive the armor might be to the motions, however, wanted to see if I could turn the character to create a reference sheet to see how it might look from multiple angles. I gave it the follow on prompt "Create a character turn around sheet for the character you just created. The sheet should include 4 angles: The front, side and Rear and three quarters (The image you just created was a three quarters - use that exact image for that angle.). Keep the character design concistent across all angles."
ChatGPT's (2025a) response :

This did change the design quite a lot, however, appeared to be referencing other comments from within the chat. Firstly it lost the armored design and balanced it with some armored elements and other more flexible parts. This worked really well and kept what I loved about some of the early responses it gave me. It has also made the nose more 'hooked' in a much more subtle way and closer to my original concept than it did when I asked it to make it more pointy. The thorns on the arms are more well balanced and subtle whilst still being a key feature of the design. The kneepads and shoulders include the gem element that also align more closely with my original design. The colours are also mostly good. I like how the green and red work together and the purple and yellow in the suit are good, however, the purple is a little too dark. The third angle from the right has lost it's Mohawk, however, the second angle works as a good side vide therefore the third angle is actually redundant. There are some odd thorns on the back that wouldn't work with my design idea of holstering the staff weapon on the back.
On the whole it was working well! I wanted to see if I could get it to generate an orthographic perspective for the back of the character because it is slightly at an angle. I gave it the follow on prompt "Draw this character so that it is a rear view but flat to the character's back rather than three quarters view. " with attached image that was cropped from the previous generation :
ChatGPT's (2025a) response :
This did mostly work, however it was still slightly turned to the right. I wanted to experiment and see if it could pose the character to help visualise how it might look in action. I gave it the follow on prompt "Show an action pose of this character, like ready to fight. Keep the design consistent with the attached design - I just want you to pose it" with the attached image cropped from a previous generation :
ChatGPT's (2025a) response :
This was fairly interesting, although the body did look a little squashed and the hand wasn't properly holding the staff. I tried another experimentation with the follow on prompt "another alternative pose"
ChatGPT's (2025a) response :
This was interesting but again felt like it was shifting the proportions ever so slightly so that they felt slightly off. At this point I felt like I'd pushed the AI to the limit and it was now at the point where I could use the generated images alognside my own artistry to create the final concept and reference sheet that I could use to model the character. It is really important that the design is strong - particularly the proportions so that when I model this it moves correctly. This was one of the key elements I wanted help with from the AI and it has supported this well. Any graphical errors in the design I can fix and rebalance through drawing now that I have a solid base.
Overall this process has been positive, but definitely wouldn't have been possible without my original sketches. This keeps the design rooted in my original ideas and concept but has helped me evolve it to a point that I am happy with the outcome. The next part in this pipeline will be to create the line art and then lock in the final colour scheme.
Final Character Line Art Development
This image shows the outcome of the character line art using ChatGPT's (2025a) responses as a base but using my own drawing skills to lock in all elements of the design I wanted to include. These tweaks such as the details around how the thorns fit into the armor, the gems fit into the shoulders, the overall symmetry and other minor details have helped create a polished design. The overall proportions achieve a strong athletic build that will work well with the type of animation I want to create. The gem on the rear will act as a holster point for the staff when this is connected.
Final Character Concept
Next I added the base colours for this design to complete it. Overall I'm really happy with the design. Whilst overall it appears sci-fi there are fantasy elements more common with medieval-fantasy such as shoulder pauldrons and boots. This creates an overall interesting aesthetic that reflect this fantasy level of technology this species has. The face works really well, I was conscious that facial rigging could be challenging but at the same time I wanted to create a character that was more alien. The mohawk, ear and nose really make this character seem more alien whilst keeping it rooted in common humanoid elements that should assist with rigging. Colours have worked particularly well with 2 complementary colour systems of green/red and purple/yellow forming the majority of the colour scheme. These are then broken up by common materials such as the brown leather books and blue/grey metal elements. The light blue in the gems also links to water, which links to the original concept of this species being gardeners and caretakers of green planets. It is like this blue water gem gives this character strength which works particularly well.

Character Weapon Concept Development
24/05/25
Next I wanted to develop the concept for the weapon the character would be using. I wanted to develop a staff that would work well as melee combat and magical ranged combat for versatility. I again used my research and preproduction materials from my previous blog post (Lawson-Hall, 2025a) to help develop these designs. I dove straight into thumbnail development to consider a range of different shapes and proportions. I iterated and reflected on each one when creating the next design - you can see I return to the ideal of the crystal being integrated into the staff numerous times. This is seen in all designs, however, the first, fifth, sixth, seventh and eighth design really iterate on this concept.
The eighth design is my most favourite because it balancing a range of sci-fi and gem-like elements to achieve a weapon that is crafted but incorporates nature. It has solid forms on each end that would work as blunt damage with a more pointy top that could blast energy as part of a ranged attack. The detailing in the middle provides a point for the character to hold or to holster on the character's back through an 'energy locking system'. Overall this design is confident enough to be taken to the next stage and inked. I've chosen not to use AI here because it felt like it wouldn't offer me anything that my own artistic skills weren't already achieving.

I created the line art through an inking process and then applied the same colours that I used in my character design to keep it consistent. Overall, its very successful with interesting forms and elements that will function as described previously. The silhouette is interesting with a range of detail and should translate into 3D easily. This completes the concepts needed for this character and next I'll combine these together to create a character sheet.
Final Character Reference Sheet
This shows the final character reference sheet together complete with the staff weapon. I'm happy with the overall presentation of this reference sheet - its clear with clean lines and the colour displays clearly. I had considered adding more detailed shading, however, I felt this wouldn't support the development of my 3D character and whilst it would improve this overall concept art it isn't necessary to my pipeline. Therefore, I left the design here so that I wasn't wasting any time on unnecessary tasks.
Further AI Exploration
Now that the final concept was complete I wanted to experiment with feeding this to the AI and seeing if it could help visualise it as a 3D model. I gave it the prompt : "Keep the proportions and the design consistent with the concept art attached. Visualise it as a 3d model" with the attached image of my concept art :
ChatGPT's (2025b) response :
"[A] Want him in a battle pose? [B] Add a weapon or magical aura? [C] Show environment (like a lava cave or forest)? [D] Cartoon version of this?"
This was interesting to see and whilst the face worked general well it has softened some features, simplified elements of the design and altered the proportions so the character looks more stocky. Regardless I thought I'd experiment with follow on prompts to see how it visualised the character. I chose option A from it's response by giving the prompt. : "A"
ChatGPT's (2025b) response :
"[B] Want to give him a flaming sword or energy weapon? [C] Add a dynamic background like a battlefield or ruined temple? [D] Want a cartoon render version too?"
This was interesting and helped me imagine what the character might look like moving. There are also some fairly interesting textures in the clothing to add an element of realism without losing the stylisation. However, again it has made the character a little more stocky than I wanted. I wanted to see if it could visualise the staff as a 3d model too so that I could imagine how it might look as a 3D model. I gave it the prompt : "Could you make this a 3d model as you have done with all the previous?" with the attached image of my concept art :
ChatGPT's (2025b) response :
"[A] Want to see this staff spinning or glowing? [B] Add animation effects like energy pulses? [C] Want it mounted on the character’s back?"
Again it had made everything a lot shorter and fatter, but I was interested to see where it might go next with it's suggestions so I responded : "Great. Option C, with the character in this chat"
ChatGPT's (2025b) response :
"[B] Want to add a glowing rune effect around the staff mount? [A] Would you like a front-facing or idle pose too? [D] Need this in a game UI concept?"
Its not quite holstered it correctly and it has changed elements of the character's design such as the armor and forearm thorns. I tried to alter the holstering effect to see if it could visualise it holstered by being locked in place by energy. I gave it the prompt : "B, like its mounted with a energy glowing effect to lock it into the Crystal on the back"
ChatGPT's (2025b) response :
"[A] Want to animate the energy lock pulsing or activating? [C] Add a combat-ready pose with energy effects? [D] Showcase in a game HUD/mockup?"
At this point it was clear that it wasn't fully understanding what I wanted or wasn't capable of visualising this. Going a long with it's options I was curious how it might look in a game environment, however, it wasn't ordering it's letters correctly - its gone with A,C and D rather than A, B and C. Therefore I prompted : "The last option" to prevent it from getting confused.
ChatGPT's (2025b) response :
"[A] Want more UI elements like mana, level, or cooldown timers? [B] Try a victory or idle animation version? [C] Need a main menu or loading screen scene?"
This was looking interesting, again the design didn't match my final concept but it was helping me see how this project could pan out. This was in a game setting but had no environment so I gave it the prompt : "Place this character in a starfox adventures inspired environment and swinging this staff around in combat"
ChatGPT's (2025b) response :
"[A] Want him facing an enemy mid-battle? [B] Add magical effects on the ground or staff? [D] Need a cinematic cutscene view?"
This was looking interesting, the pose was very off balance and the design had changed slightly, however, it was good to imagine this character in a world. I felt like I was reaching the end of it's capabilities and wanted to try one last experiment from it's suggestions. I gave the prompt : "A."
ChatGPT's (2025b) response :
"[B] Want lightning or magical trails from the staff swing? [C] Add a jungle enemy facing him? [D] Turn this scene into a poster or splash screen?"
I then had a splinter idea to see if it could create a turn around sheet of my character but from the 3D model to compare to my concept. I gave it the prompt : "Can you make a top/side/back reference sheet to assist with modeling?"
ChatGPT's (2025b) response :
This was pretty interesting because it has turned the staff slightly too. However, when you inspect closer so many elements of the design aren't consistent such as the detailing on the top of the staff, the rear of the character has unwanted elements, however, it is good to visualise what it might look like in 3d. In particular, the side view of the head almost perfectly translates the hooked nose into 3 dimensions to help me visualise how the geometry might connect together with the cheeks and other elements. The mohawk across all three has a good chunky-ness to it that will help when giving this volume. The indentations on the staff are good and help show how this could be modelled to look solid when it is modelled. Furthermore, it shows where some creases might fall such as around the brow and muscles.
Despite this as mentioned throughout this blog, none of this is production ready and serves as a visualisation tool to look into the future. This should be used in combination with traditional research. For example, whilst this shows how the face might be structured in 3D I will also be studying anatomy to really understand how muscles should sit underneath the skin to give this athletic physique. I will also need to refer to my concept art to accurately design the space ranger suit since this morphs and changes slightly across these images. However, prior to AI I wouldn't have been able to create these visualisations and it has really enhanced the production process to confirm that the design is working and should translate well into 3D.
Conclusion
In conclusion, this has been a highly successful process of designing a character and producing a reference sheet for modelling. This is such a crucial step in the process to perfect because it is laying the foundations for your character production that will ultimately be animated. Any problems that are baked in here could cause issues later down the production line. I'm particularly happy with how in-depth I explored the character profile and other elements linked to their worldbuilding in my previous blog post (Lawson-Hall, 2025a). This laid a really strong foundation that has helped this character stand out and feel real. There are just enough plant-like and alien elements to make them feel like another species without staying too far from a humanoid base that it might be challenging to represent realistic body mechanics and facial animation.
AI has been an extremely powerful tool to experiment with in this part of the production pipeline. I strongly believe that if I hadn't have explored the character's worldbuilding and initial designs then I would have ended up with a sub par result. Feeding the AI my sketches and tweaking through prompts has enabled me to accurately refine proportions, simplify elements of the design and iterate through concept whilst still staying true to my original artist direction. It has also been a powerful tool to look forward into the pipeline to see how this design might translate into 3D. I am confident that a trained concept artist would likely have performed better, however, I'm focussed on developing skills in animation, therefore using AI to assist in skills I am less proficient in has been a powerful tool to ensure that the character is well designed and ready for the next stage in the production pipeline.
Project Management Reflection
Overall the project is now at a good point to progress onto 3D modelling. The image below shows the completed Trello card. This was completed slightly later than planned on the 24th May rather than the 19th, however, I have taken a little bit of time off because technically the next module hasn't started yet. I'm waiting for lecturers to confirm the start date, however they have indicated it would be around the end of May/beginning of June. Therefore, some rest and relaxation before going full steam into this project will be good to rest and recharge - whilst still making some progress.
This was the production schedule I created and I am 1 week behind, however, I think I may have been slightly ambitious. Also there is a lot of time in the later parts of the project that I could readjust so that I'm focussing on perfecting each stage of the production pipeline. Also this is to be expected as I mentioned I slowed up a little to rest and recharge. I am planning on tweaking this to give me a little more time to complete the elements I need to a high standard.
I've tweaked the production schedule slightly to allow for a little more time modelling and texturing and then shunted everything over by 2 weeks. This should give me enough time to establish a strong character that can be then used for rigging and ultimately animation. I also have 4 weeks in summer where I will be able to work on this full time and a week in October. This will allow for flexibility in the production pipeline as well.
Reference List
- ChatGPT (2025a) ChatGPT response to Matt Lawson-Hall, 13th May.
- ChatGPT (2025b) ChatGPT response to Matt Lawson-Hall, 24th May.
- Lawson-Hall, M. (2025a) Preproduction Exploration to Prepare for Career Development in the Graduate Portfolio Module. Matt Lawson-Hall Animation Blog. Available at: https://mattlawsonhall.blogspot.com/2025/04/practical-skill-development-to-prepare.html (Accessed: 27/05/25).
- Lawson-Hall, M. (2025b) AI's Not-So-Secret Invasion: Disruption in the Screen Industries: Is AI a Catalyst for Change or a Cause for Concern? Matt Lawson-Hall Animation Blog. Available at: https://mattlawsonhall.blogspot.com/2025/04/ais-not-so-secret-invasion-disruption.html (Accessed: 27/05/25).