/ Blogs & Newsletters
Sogni Labs · Image generation

MOSS Image LoRAs: Fever Dreams, Kodachrome and Four More

Give your MOSS Image creations a new mood with six community LoRAs: small add-ons that guide the model’s style and detail. Discover dreamy portraits, vintage photos and more, with side-by-side examples and prompts you can try in Sogni.

Prompt · seed 880103 · 1536×1920

Top-down close-up from directly above, face large and sharp in the center. A woman with dark wet hair spread around her, natural skin texture with visible pores and fine freckles, sharp clear eyes with natural catchlights, wearing a black velvet long-sleeved gown, lies in a bathtub of milk with floating pink rose petals, looking up into the lens. Soft window light.

LoRAs: moss-fever-dream-mood 1.0 · moss-mystic-x 0.8 · both stacked at 0.8 + 0.7

Milk bath, from above: No LoRA
No LoRAoff
Milk bath, from above: Fever Dream
Fever Dream1.0
Milk bath, from above: Mystic X
Mystic X0.8
Milk bath, from above: Both
Both0.8 + 0.7
One prompt, one seed, four different pictures. Frame 03 of six. MOSS Image, 50 steps, seed 880103. Only the attached LoRAs changed across these four renders. Explore this example ↓

MOSS Image now has eleven community LoRAs to explore on Sogni. The six new arrivals give you more ways to set the scene: a faded family photo, a surreal portrait or the look of a camcorder recording.

The short version

For a film-inspired look, start with Kodachrome Photo or Retro Vintage Photo. Try Fever Dream Mood for glow and surreal color, or Mystic X for skin and material detail. 2000s Analog Core brings camcorder nostalgia, while MAX ULTRA hires explores a more detailed finish. Choose one, adjust its strength, or combine two to find your own look.

Get started

Three ways in

Start in the browser, ask Sogni Chat, or use your own creative agent. All eleven LoRAs work with MOSS Image on the Sogni Supernet.

01

In the browser

Open Sogni, choose MOSS Image as the model, open the LoRA panel and pick a look. Drag the slider to set the strength.

app.sogni.ai/create ↗
02

In Sogni Chat

Describe the picture and name the LoRA you want, for example “MOSS Image with Fever Dream Mood at 1.0”. The assistant sets up the render for you.

chat.sogni.ai ↗
03

In your own agent

Install the Sogni Creative Agent Skill in Claude Code, Codex CLI or Hermes and render straight from the terminal, in batches if you like.

sogni.ai/agent ↗
Using your own agent? Copy the setup and render commands
install: sogni creative agent skill
Claude Code
/plugin marketplace add Sogni-AI/sogni-creative-agent-skill /plugin install sogni-creative-agent@sognirun inside Claude Code
Render
sogni-agent -m moss_image_sft_int8_convrot --lora moss-fever-dream-mood --lora-strength 1.0 -w 1536 -h 1920 "a woman in an emerald velvet gown in a bathtub full of pink peonies"any LoRA id from the shelf below works the same way
List
sogni-agent --list-loras --lora-catalog-model moss_image_sft_int8_convrotevery MOSS LoRA with its id, range and default
Start here

What changes when you add a LoRA?

A LoRA guides how the model creates a new image. Along with color and texture, it can change the pose, framing and subject, even when you keep the same prompt and seed.

The six new ones

Choose your starting look

Kodachrome Photo by lotrab921 recalls 1960s and 1970s color slides, and it is trained around the trigger kodachrome style. Retro Vintage Photo by patrik goes for found photographs and family albums from the late fifties to the early nineties. Both have a strength range of 0 to 1.2 and default to 1.

Fever Dream Mood, also by patrik, adds dreamy analog color, luminous highlights and a surreal edge; it defaults to 0.8. Mystic X by alcaitiff emphasizes skin, anatomy and material detail, with support for mature themes. It defaults to 1; the portraits below show it at 0.8.

2000s Analog Core by Danrisi brings early-2000s VHS and Hi8 camcorder color and exposure. MAX ULTRA hires by comfyanonymous is a high-resolution adapter designed for canvases of 4 MP and above. Frames 05 and 06 put them against each other.

How to read the rows

Six scenes, four renders each

Each row starts with MOSS Image alone, followed by one LoRA, the other LoRA, and both together. The prompt, seed, size (1536×1920) and 50-step setting stay the same within each row. The labels show each LoRA’s strength.

Look at the light, color, texture and framing, then copy a prompt to try a look yourself.

Frame 01

Desert motel

Prompt · seed 880101 · 1536×1920

kodachrome style, portrait from the waist up, face large and sharp. A young woman in a yellow sundress and a white silk headscarf, natural skin texture with visible pores and fine freckles, sharp clear eyes with natural catchlights, sits on the hood of a turquoise 1950s convertible outside a desert motel, smiling into the lens, a pink neon motel sign and blue sky behind her. Late afternoon sun.

LoRAs: moss-kodachrome-photo 1.1 · moss-retro-vintage-photo 1.1 · both stacked at 0.8 + 0.8

Desert motel: No LoRA
No LoRAoff
Desert motel: Kodachrome
Kodachrome1.1
Desert motel: Retro Vintage
Retro Vintage1.1
Desert motel: Both
Both0.8 + 0.8

In this scene, Kodachrome changes the framing and hairstyle along with the color. Retro Vintage stays closer to the original composition, with a faded photographic look. The combined version brings the period styling into the mix.

Frame 02

The fisherman

Prompt · seed 880102 · 1536×1920

kodachrome style, close-up portrait from the chest up, face large and sharp. An elderly fisherman with a white beard, natural skin texture with visible pores and fine freckles, sharp clear eyes with natural catchlights, wearing a navy knitted cap and a yellow oilskin coat, at a small harbor, red and blue wooden boats behind him, squinting kindly into the lens. Bright morning light.

LoRAs: moss-kodachrome-photo 1.1 · moss-retro-vintage-photo 1.1 · both stacked at 0.8 + 0.8

The fisherman: No LoRA
No LoRAoff
The fisherman: Kodachrome
Kodachrome1.1
The fisherman: Retro Vintage
Retro Vintage1.1
The fisherman: Both
Both0.8 + 0.8

Kodachrome gives this portrait warmer skin tones and a softer background. Retro Vintage brings out a more weathered look. Combining them gives the harbor scene a richer color palette.

Frame 03

Milk bath, from above

Prompt · seed 880103 · 1536×1920

Top-down close-up from directly above, face large and sharp in the center. A woman with dark wet hair spread around her, natural skin texture with visible pores and fine freckles, sharp clear eyes with natural catchlights, wearing a black velvet long-sleeved gown, lies in a bathtub of milk with floating pink rose petals, looking up into the lens. Soft window light.

LoRAs: moss-fever-dream-mood 1.0 · moss-mystic-x 0.8 · both stacked at 0.8 + 0.7

Milk bath, from above: No LoRA
No LoRAoff
Milk bath, from above: Fever Dream
Fever Dream1.0
Milk bath, from above: Mystic X
Mystic X0.8
Milk bath, from above: Both
Both0.8 + 0.7

Fever Dream adds a glossy glow to the skin and water. Mystic X stays closer to the original portrait. The combined version keeps the glow with a softer finish.

Frame 04

Moonlit greenhouse

Prompt · seed 880104 · 1536×1920

Portrait from the chest up, face large and sharp. A woman with a dark low chignon and pearl earrings, natural skin texture with visible pores and fine freckles, sharp clear eyes with natural catchlights, deep red lips, wearing a red silk dress with long sleeves, stands in a moonlit Victorian greenhouse full of white orchids and low fog, looking over her shoulder into the lens.

LoRAs: moss-fever-dream-mood 1.0 · moss-mystic-x 0.8 · both stacked at 0.8 + 0.7

Moonlit greenhouse: No LoRA
No LoRAoff
Moonlit greenhouse: Fever Dream
Fever Dream1.0
Moonlit greenhouse: Mystic X
Mystic X0.8
Moonlit greenhouse: Both
Both0.8 + 0.7

Both LoRAs brighten this moonlit portrait and bring out skin texture. Fever Dream gives the scene cooler light and changes the dress, while Mystic X stays closer to the original framing.

Frame 05

Snow on the roof

Prompt · seed 880105 · 1536×1920

Portrait from the chest up, face large and sharp. A teenage girl with butterfly clips in her dark hair, natural skin texture with visible pores and fine freckles, sharp clear eyes with natural catchlights, glossy lips, wearing a silver puffer jacket, stands on a snowy city rooftop at night, snowflakes in the air, city lights and a water tower behind her, grinning into the lens.

LoRAs: moss-2000s-analog-core 1.0 · moss-max-ultra-hires 1.0 · both stacked at 0.8 + 0.8

Snow on the roof: No LoRA
No LoRAoff
Snow on the roof: 2000s Analog
2000s Analog1.0
Snow on the roof: MAX ULTRA
MAX ULTRA1.0
Snow on the roof: Both
Both0.8 + 0.8

2000s Analog Core changes the hairstyle and gives the portrait a stronger flash look. MAX ULTRA hires produces a darker image with deeper color. The combined version brings another balance of light and color.

Frame 06

Thrift shop cat

Prompt · seed 880106 · 1536×1920

A ginger cat asleep on a tall pile of colorful vintage knitted sweaters in a cluttered thrift shop, racks of clothes, old lamps, stacked records and a dusty window behind. Warm afternoon light, every knit stitch and hair visible.

LoRAs: moss-2000s-analog-core 1.0 · moss-max-ultra-hires 1.0 · both stacked at 0.8 + 0.8

Thrift shop cat: No LoRA
No LoRAoff
Thrift shop cat: 2000s Analog
2000s Analog1.0
Thrift shop cat: MAX ULTRA
MAX ULTRA1.0
Thrift shop cat: Both
Both0.8 + 0.8

2000s Analog Core gives the shop warmer light and changes the knitwear. MAX ULTRA hires produces a darker scene. Compare the detail crops below to explore the sweaters and fur.

A closer look at the details

Compare the center of each image in these 420×420 detail crops. Look at the knit stitches, fur and lighting.

No LoRA detail crop
No LoRA
2000s Analog detail crop
2000s Analog
MAX ULTRA detail crop
MAX ULTRA
Both detail crop
Both
Strength, step by step

How much is enough?

Strength controls how strongly a LoRA guides the result. Each row shows the same prompt and seed at several strengths, alongside a version with no LoRA. On mobile, swipe across a row to compare the settings. Start with the default, then move the slider to find a look you like.

Fever Dream Mood

Glow, blush and grain build steadily; by 1.2 the whole frame is soft-focus memory.

Fever Dream Mood at No LoRA
No LoRA
Fever Dream Mood at 0.4
0.4
Fever Dream Mood at 0.8
0.8
Fever Dream Mood at 1.2
1.2

Mystic X

Subtle on a portrait this restrained: skin and hair pick up texture and contrast while the face barely moves.

Mystic X at No LoRA
No LoRA
Mystic X at 0.4
0.4
Mystic X at 0.8
0.8
Mystic X at 1.2
1.2

Kodachrome Photo

Hair, makeup and color take on an early-sixties look as the strength increases.

Kodachrome Photo at No LoRA
No LoRA
Kodachrome Photo at 0.4
0.4
Kodachrome Photo at 0.8
0.8
Kodachrome Photo at 1.2
1.2

Retro Vintage Photo

Color warms and fades, and the photo starts to read as a 1970s print. It also nudges the age and styling of the subject.

Retro Vintage Photo at No LoRA
No LoRA
Retro Vintage Photo at 0.4
0.4
Retro Vintage Photo at 0.8
0.8
Retro Vintage Photo at 1.2
1.2

2000s Analog Core

Subtle at this framing: the flash gets harsher and the color cooler.

2000s Analog Core at No LoRA
No LoRA
2000s Analog Core at 0.4
0.4
2000s Analog Core at 0.7
0.7
2000s Analog Core at 1
1

MAX ULTRA hires

Compare the texture of moss, rock and spray. These examples use a 2.4 MP canvas; the creator designed this adapter for 4 MP and above.

MAX ULTRA hires at No LoRA
No LoRA
MAX ULTRA hires at 0.4
0.4
MAX ULTRA hires at 0.8
0.8
MAX ULTRA hires at 1.2
1.2

Canon UltraReal

The background melts into creamy blur and the portrait picks up glossier, editorial color.

Canon UltraReal at No LoRA
No LoRA
Canon UltraReal at 0.4
0.4
Canon UltraReal at 0.7
0.7
Canon UltraReal at 1
1

Lenovo UltraReal

Light flattens into real window light and the frame starts to feel like a phone photo. From 0.75 it also nudges the pose.

Lenovo UltraReal at No LoRA
No LoRA
Lenovo UltraReal at 0.25
0.25
Lenovo UltraReal at 0.5
0.5
Lenovo UltraReal at 0.75
0.75
Lenovo UltraReal at 1.0
1.0

GrainScape UltraReal

Contrast and texture climb, the fog gets moodier and the color more filmic.

GrainScape UltraReal at No LoRA
No LoRA
GrainScape UltraReal at 0.4
0.4
GrainScape UltraReal at 0.7
0.7
GrainScape UltraReal at 1
1

NiceGirls UltraReal

A gentle portrait polish: smoother skin and a slightly warmer, more styled look.

NiceGirls UltraReal at No LoRA
No LoRA
NiceGirls UltraReal at 0.4
0.4
NiceGirls UltraReal at 0.7
0.7
NiceGirls UltraReal at 1
1

Detail Tweaker

At −0.75 the scene becomes simpler; at +0.75 the shelves and skin gain detail.

Detail Tweaker at -0.75
-0.75
Detail Tweaker at No LoRA
No LoRA
Detail Tweaker at 0.4
0.4
Detail Tweaker at 0.75
0.75
The whole shelf

Every LoRA, one prompt

All eleven against a single café scene at 2048×1152, seed 777001. Kodachrome, Canon and Lenovo get their trigger word at the start of the prompt.

No LoRA
No LoRAoff
Fever Dream Mood
Fever Dream Mood1.0
Mystic X
Mystic X0.8
Kodachrome Photo
Kodachrome Photo1.1
Retro Vintage Photo
Retro Vintage Photo1.1
2000s Analog Core
2000s Analog Core1.0
MAX ULTRA hires
MAX ULTRA hires1.0
Canon UltraReal
Canon UltraReal1.0
Lenovo UltraReal
Lenovo UltraReal1.0
GrainScape UltraReal
GrainScape UltraReal1.0
NiceGirls UltraReal
NiceGirls UltraReal1.0
Detail Tweaker
Detail Tweaker0.75
The full catalogue

The MOSS shelf

Each card shows the range the slider offers in the app, with the recommended band in mint and the default marked. Creator links open the original model pages.

Looks and eras

6 LoRAs · new

Explore a period, a medium or a mood. Each LoRA can also influence the subject and composition.

Fever Dream MoodNew

patrik

0+1.2

Default: 0.8 · Suggested: 0.8 to 1.2

Positive only

moss-fever-dream-mood

Mystic XNew

alcaitiff

0+1.2

Default: 1 · Suggested: 1 to 1.2

Positive only

Mature

moss-mystic-x

Kodachrome PhotoNew

lotrab921

0+1.2

Default: 1 · Suggested: 1 to 1.2

Positive only

Trigger: kodachrome style

moss-kodachrome-photo

Retro Vintage PhotoNew

patrik

0+1.2

Default: 1 · Suggested: 1 to 1.2

Positive only

moss-retro-vintage-photo

2000s Analog CoreNew

Danrisi

0+1

Default: 0.7 · Suggested: 0.7 to 1

Positive only

moss-2000s-analog-core

MAX ULTRA hiresNew

comfyanonymous

0+1.2

Default: 1 · Suggested: 1 to 1.2

Positive only

moss-max-ultra-hires

Photographic realism

4 LoRAs

Camera and film looks. Canon and Lenovo are trained around a trigger word; put it at the start of the prompt.

Canon UltraReal

Danrisi

0+1

Default: 0.7 · Suggested: 0.7 to 1

Positive only

Trigger: c2n0n

moss-canon-ultrareal

Lenovo UltraReal

Danrisi

0+1

Default: 0.7 · Suggested: 0.7 to 1

Positive only

Trigger: l3n0v0

moss-lenovo-ultrareal

GrainScape UltraReal

Danrisi

0+1

Default: 0.7 · Suggested: 0.7 to 1

Positive only

moss-grainscape-ultrareal

NiceGirls UltraReal

Danrisi

0+1

Default: 0.6 · Suggested: 0.6 to 1

Positive only

moss-nicegirls-ultrareal

Detail

1 LoRA

Positive values add detail; negative values simplify it. This experimental adapter was trained on anime.

Detail Tweaker

lse14

-0.75+0.75

Default: 0.6 · Suggested: -0.75 to 0.75

SimplerMore detail

moss-detail-tweaker

Your own look

Bring your own LoRA

The shelf is a starting point, not a limit. If you have a MOSS Image LoRA of your own, or one from a creator you like, you can import it as a Personal LoRA and use it exactly like the eleven above, stacked with them if you want. Paste a Hugging Face .safetensors link or a Civitai model link, give it a name, and once it has been validated it shows up in your account next to the shared shelf.

Importing a LoRA does not train anything new; it brings an existing adapter onto the Sogni Supernet. You confirm that you have the rights to use the file when you import it. In the browser, open the LoRA panel and choose My LoRAs to get started.

Import and use a Personal LoRA from your agent
personal lora: import and render
Import
sogni-agent --import-lora https://huggingface.co/author/repo/resolve/main/style.safetensors --personal-lora-name "My look" --personal-lora-model moss_image_sft_int8_convrot --confirm-lora-rightsHugging Face .safetensors or Civitai model links
Check
sogni-agent --list-personal-lorasready imports can render
Render
sogni-agent -m moss_image_sft_int8_convrot --lora personal-YOUR-ID --lora moss-kodachrome-photo --lora-strength 0.8 --lora-strength 0.8 "your prompt"stack your own LoRA with any from the shelf
Practical notes

Working with them

Creators and sources

  1. patrik: Fever Dream Mood
  2. patrik: Retro Vintage Photo
  3. alcaitiff: Mystic X
  4. lotrab921: Kodachrome Photo
  5. Danrisi: 2000s Analog Core, plus Canon, Lenovo, GrainScape and NiceGirls UltraReal
  6. comfyanonymous: MOSS Image MAX ULTRA hires
  7. lse14: MOSS detail tweaker
  8. Sogni Labs: Krea 2 Community LoRAs

Comparison settings: MOSS Image Standard SFT, 50 steps, guidance (CFG) 3, 1536×1920, seeds 880101 to 880106. Strength examples use 1152×2048; the café comparison uses 2048×1152.

Make the look your own

Pick a LoRA, copy a prompt and explore your next image. Start with a single look, then adjust the strength or try a combination.