MOSS Serenity
MOSS Serenity is a full finetune of MOSS-Image Base that teaches the model a set of people by name, while keeping what the base model already knows. The current release, v6, was trained for 210,000 steps on a single RTX 5090 with maltrainer, in seven stages that each added more people (Versions).
It is a drop-in replacement for MOSS-Image Base: the same architecture, text encoder (Qwen3.5-9B) and VAE (FLUX.2), so any MOSS-Image workflow in ComfyUI runs it unchanged.
 |
 |
 |
 |
 |
 |
 |
 |
The v6 gallery shows all 179 newest samples (faces, showcases, styles) with their prompts and shared settings.
More in samples/: v6 (faces, showcases, styles), v5 and v4 (film portraits of earlier models), v3 and v2 (portraits and 50-step scenes), portraits (8-step street portraits),
scenes (8-step film scenes), scenes-50steps (Full HD, 50 steps),
pairs (two people), trios (three people) and no-people.
Files
| File | What |
|---|---|
serenity_base_v6_210000_int8_convrot.safetensors |
latest: v6 in ComfyUI's INT8 ConvRot format (5.4 GB): half the size and faster, within ~1% of the bf16 weights |
serenity_base_v5_170000_bf16.safetensors |
the newest bf16: the finetuned DiT, bf16, ComfyUI single file (10.2 GB): the EMA average of the weights at step 170,000 |
serenity_base_v3_100000_int8_convrot.safetensors |
v3 in ComfyUI's INT8 ConvRot format (5.4 GB): half the size and faster, within ~1% of the bf16 weights |
serenity_base_v1_24000_int8_convrot.safetensors |
v1 in INT8 ConvRot |
lora-extract/serenity_v<N>_lora_r64.safetensors |
each version (v1-v6) as a LoRA for MOSS-Image Base, rank 64 (0.3 GB) |
lora-extract/serenity_v<N>_lora_r128.safetensors |
rank 128 (0.6 GB) |
lora-extract/serenity_v<N>_lora_r256.safetensors |
rank 256 (1.2 GB) |
workflow/base-turbo-no-lora.json |
ComfyUI workflow: 8 steps with the DMD turbo LoRA |
workflow/base-no-turbo-no-lora.json |
ComfyUI workflow: 50 steps, no turbo |
workflow/base-turbo-lora.json |
ComfyUI workflow: turbo, plus a person LoRA of your own in the Power Lora Loader |
workflow/lora-extract-turbo-lora.json |
ComfyUI workflow: the original MOSS-Image Base + a Serenity LoRA (the file names serenity_v1_lora_r256; pick any version in the loader) + turbo, 12 steps |
docs/generating.md |
generating with the model: setup, settings, prompting |
docs/training.md |
how it was trained, and how to train or continue it |
The text encoder and VAE are MOSS-Image's own: Qwen3.5-9B (qwen3.5_9b_int8_convrot.safetensors,
or qwen3.5_9b_bf16.safetensors)
in models/text_encoders, and the FLUX.2 VAE (flux2_vae.safetensors) in models/vae.
Versions
Each version continues the previous one from its saved state (weights, optimizer and EMA), with more people added. The people in the Characters table are the ones trained up to v6.
| Version | Steps | File here |
|---|---|---|
| v1 | 24,000 | INT8 ConvRot, LoRAs |
| v2 | 60,000 | LoRAs |
| v3 | 100,000 | INT8 ConvRot, LoRAs |
| v4 | 135,000 | LoRAs |
| v5 | 170,000 | bf16, LoRAs |
| v6 | 210,000 | INT8 ConvRot, LoRAs |
Later versions know everyone earlier ones did, but a person trained long ago and rehearsed little since can be weaker in a later version than in the one where they were new: compare two versions on the same seed if one person matters to you (samples/v4, samples/v5).
Quick start (ComfyUI)
- Put
serenity_base_v6_210000_int8_convrot.safetensors(the latest, half the size and faster; orserenity_base_v5_170000_bf16.safetensorsfor bf16) inComfyUI/models/diffusion_models/. - Drag one of the workflows into ComfyUI and pick the model in Load Diffusion Model.
- Prompt
photo of <trigger>, a woman, ...(see Prompting).
| Turbo (fast) | Full (best detail) | |
|---|---|---|
| Extra LoRA | lodestones' MOSS-Image-Base-8-steps-dmd at 1.0 |
none |
| Steps | 8 | 50 (20 is close for landscapes) |
| CFG | 4.0 | 4.0 |
| Sampler / scheduler | euler / simple | euler / simple |
| SkipLayerGuidanceDiTSimple | layer 22, start 0, end 0.5 | layer 22, start 0, end 0.5 |
| Negative | bad ugly blurry |
bad ugly blurry |
| Size used for the samples | 1584 ร 1776 | 1920 ร 1088 |
| Time on an RTX 5090 | ~16 s |
Shift needs no node: ComfyUI runs MOSS-Image at shift 5. Full details in docs/generating.md.
bf16 or INT8 ConvRot? Same model, same workflows: pick the file in Load Diffusion Model. The
INT8 file stores each block's big Linears as int8 in a Hadamard-rotated basis with one scale per
row (ComfyUI's INT8 ConvRot, as in Comfy-Org's moss_image_*_int8_convrot releases), so it is half
the size and runs on comfy-kitchen's int8 kernels. Its weights are within 0.9% of the bf16 ones
(mean relative error per Linear), and on the same seed it renders a near-identical image: the turbo
workflow took 8.7 s including loading, against ~16 s for bf16, on an RTX 5090. Keep the bf16 file
for finetuning, merging and LoRA extraction.
Prompting
The model learned each person from captions of one shape, so prompts work best in that shape:
photo of <trigger>, a <woman|man>, <one sentence: look, framing, clothing, setting, light>
photo of gilliananderson, a woman with shoulder-length straight auburn red hair, a serious focused
expression, medium close-up from the chest up, holding a small flashlight raised beside her face,
wearing a dark charcoal 1990s pantsuit jacket over a white blouse, standing in a dark abandoned
warehouse corridor at night, cold blue moonlight, 1990s film grain, cinematic film still
- Keep faces large. Chest-up and closer gives the best likeness; full-body shots with a small face can come out wrong.
- Describe the look, not the title. The model knows the people, not their films: "a long brown leather duster coat in the cargo bay of a worn frontier spaceship" works, a film title does not.
- Two people: a man and a woman work well; two of the same gender are hit-or-miss. See Two people in one shot and Three people.
- Without a trigger,
photo of a woman, ...gives an ordinary person, as in the base model.
Characters
Each trained person is prompted with a trigger: their name in lower case, without spaces, put
first after photo of, then the class word they were trained with:
photo of keanureeves, a man, ...
The table lists everyone the model was trained on up to v6 (210,000 steps). Samples marked v6, v5 and v4 are film portraits
from those models; v2 and v3 samples come from the earlier versions; a - means there is no sample in this repository yet.
The samples below were made with the workflows in workflow/: (50) marks the full
model at 50 steps and 1920 ร 1088; the rest are the 8-step turbo workflow at 1584 ร 1776.
| Trigger | Class | Who | Samples |
|---|---|---|---|
aleksandrasawicka |
woman | Aleksandra Sawicka, Instagram/TikTok | film portrait (v5), street portrait (v2), film scene (v2, 50) |
alexandramarwasaber |
woman | Denims (Alexandra Marwasaber), Twitch streamer | - |
alisonbrie |
woman | Alison Brie | film portrait (v5), street portrait, Community, GLOW, GLOW (50) |
amberrayne |
woman | Amber Rayne | film portrait (v5), film portrait (v4), film noir street (50) |
angelinajolie |
woman | Angelina Jolie | - |
anyachalotra |
woman | Anya Chalotra | film portrait (v5), film portrait (v4), street portrait, The Witcher (50) |
anyataylorjoy |
woman | Anya Taylor-Joy | film portrait (v5), film portrait (v4), street portrait, The Queen's Gambit, The Queen's Gambit (50) |
asmongold |
man | Asmongold, streamer | film portrait (v5), film portrait (v4), streaming (50), stream reaction |
aubreyplaza |
woman | Aubrey Plaza | film portrait (v6), film portrait (v5), film portrait (v4), street portrait, The White Lotus, The White Lotus (50) |
beniciodeltoro |
man | Benicio del Toro | street portrait (v2), film scene (v2, 50) |
billieeilish |
woman | Billie Eilish | film portrait (v5), film portrait (v4), street portrait (v2), film scene (v2, 50) |
bradpitt |
man | Brad Pitt | film portrait (v5), film portrait (v4), street portrait, Troy (50) |
brielarson |
woman | Brie Larson | street portrait (v2), film scene (v2, 50) |
britneyspears |
woman | Britney Spears | film portrait (v4), street portrait (v2), film scene (v2, 50) |
caitylotz |
woman | Caity Lotz | film portrait (v4) |
camerondiaz |
woman | Cameron Diaz | - |
camrenbicondova |
woman | Camren Bicondova | street portrait (v2), film scene (v2, 50) |
caradelevingne |
woman | Cara Delevingne | film portrait (v4) |
caragee |
woman | Cara Gee | - |
carlagugino |
woman | Carla Gugino | - |
cateblanchett |
woman | Cate Blanchett | film portrait (v4), street portrait (v2), film scene (v2, 50) |
cathrinebell |
woman | Catherine Bell, actress | film portrait (v4) |
cbg |
woman | Comic Book Girl 19 (CBG), YouTuber | - |
charlizetheron |
woman | Charlize Theron | film portrait (v5), film portrait (v4), street portrait (v2), film scene (v2, 50) |
chrishemsworth |
man | Chris Hemsworth | street portrait (v2), film scene (v2, 50) |
chrisrock |
man | Chris Rock | film portrait (v5), film portrait (v4) |
clairedanes |
woman | Claire Danes | street portrait (v2), film scene (v2, 50) |
colinfarrell |
man | Colin Farrell | street portrait (v2), film scene (v2, 50) |
dafnekeen |
woman | Dafne Keen | film portrait (v5) |
danielcraig |
man | Daniel Craig | film portrait (v6), film portrait (v5), film portrait (v4), street portrait (v2), film scene (v2, 50) |
dariadabrowska |
woman | Daria Dฤ browska, Instagram/TikTok | film portrait (v5), film portrait (v4) |
davidduchovny |
man | David Duchovny | film portrait (v5), film portrait (v4), street portrait (v2), film scene (v2, 50) |
davidtennant |
man | David Tennant | film portrait (v5) |
devonaoki |
woman | Devon Aoki | film portrait (v5) |
donnabae |
woman | Doona Bae, actress | film portrait (v6), film portrait (v4) |
donnieyen |
man | Donnie Yen | film portrait (v4) |
edwardnorton |
man | Edward Norton | film portrait (v6), film portrait (v5), film portrait (v4) |
elizabethdebicki |
woman | Elizabeth Debicki | film portrait (v4) |
elizabethmitchell |
woman | Elizabeth Mitchell | street portrait (v2), film scene (v2, 50) |
elizabetholsen |
woman | Elizabeth Olsen | film portrait (v5), street portrait (v2), film scene (v2, 50) |
ellenpage |
woman | Elliot Page, pre-transition photos | - |
elliekemper |
woman | Ellie Kemper | film portrait (v5), film portrait (v4) |
eminem |
man | Eminem | film portrait (v5) |
emmastone |
woman | Emma Stone | film portrait (v6), film portrait (v5), film portrait (v4), street portrait (v2), film scene (v2, 50) |
emmawatson |
woman | Emma Watson | film portrait (v5), film portrait (v4), Beauty and the Beast ballroom, Hogwarts library (50) |
erinmoriarty |
woman | Erin Moriarty | film portrait (v5), street portrait (v2), film scene (v2, 50) |
evagreen |
woman | Eva Green | film portrait (v5), film portrait (v4), street portrait (v2), film scene (v2, 50) |
evangelinelilly |
woman | Evangeline Lilly | street portrait (v2), film scene (v2, 50) |
evanrachelwood |
woman | Evan Rachel Wood | - |
ewaswoboda |
woman | Ewa Swoboda, sprinter | - |
feliciaday |
woman | Felicia Day | film portrait (v5), film portrait (v4), street portrait, The Guild, The Guild (50) |
felicityjones |
woman | Felicity Jones | film portrait (v5), film portrait (v4), street portrait (v2), film scene (v2, 50) |
florencepugh |
woman | Florence Pugh | film portrait (v6), film portrait (v5), film portrait (v4) |
freyaallan |
woman | Freya Allan | film portrait (v5), film portrait (v4) |
galgadot |
woman | Gal Gadot | film portrait (v6), film portrait (v5), film portrait (v4), street portrait (v2), film scene (v2, 50) |
gilliananderson |
woman | Gillian Anderson | film portrait (v6), film portrait (v5), film portrait (v4), street portrait, The X-Files (50) |
ginacarano |
woman | Gina Carano | street portrait (v2), film scene (v2, 50) |
ginatorres |
woman | Gina Torres | film portrait (v5), film portrait (v4) |
gracerandolph |
woman | Grace Randolph | - |
guypearce |
man | Guy Pearce | film portrait (v4), street portrait (v2), film scene (v2, 50) |
gwynethpaltrow |
woman | Gwyneth Paltrow | film portrait (v6), film portrait (v5), film portrait (v4) |
haileesteinfeld |
woman | Hailee Steinfeld | film portrait (v5), film portrait (v4), street portrait (v2), film scene (v2, 50) |
halleberry |
woman | Halle Berry (trigger halleberry2 in v5 and earlier) |
- |
hayleyatwell |
woman | Hayley Atwell | street portrait (v2), film scene (v2, 50) |
helenhunt |
woman | Helen Hunt | film portrait (v4), street portrait (v2), film scene (v2, 50) |
helenmirren |
woman | Helen Mirren | street portrait, RED, RED (50) |
idriselba |
man | Idris Elba | street portrait (v2), film scene (v2, 50) |
jamiefoxx |
man | Jamie Foxx | - |
jamieleecurtis |
woman | Jamie Lee Curtis | - |
januszkorwinmikke |
man | Janusz Korwin-Mikke | film portrait (v5), film portrait (v4), street portrait (v2), film scene (v2, 50) |
jasonmomoa |
man | Jason Momoa | film portrait (v5), Aquaman, Aquaman (50), street portrait |
javierbardem |
man | Javier Bardem | film portrait (v6), film portrait (v5), film portrait (v4), street portrait (v2), film scene (v2, 50) |
jennaortega |
woman | Jenna Ortega | Wednesday (50) |
jenniferaniston |
woman | Jennifer Aniston | film portrait (v4), street portrait (v2), film scene (v2, 50) |
jennifercarpenter |
woman | Jennifer Carpenter | film portrait (v5), film portrait (v4), street portrait (v2), film scene (v2, 50) |
jenniferholland |
woman | Jennifer Holland | - |
jenniferlawrence |
woman | Jennifer Lawrence | film portrait (v5), film portrait (v4), The Hunger Games, The Hunger Games (50), Silver Linings Playbook |
jenniferlopez |
woman | Jennifer Lopez | film portrait (v5) |
jennifertilly |
woman | Jennifer Tilly | - |
jenpsaki |
woman | Jen Psaki | film portrait (v6), film portrait (v5), film portrait (v4), street portrait (v2), film scene (v2, 50) |
joannakulig |
woman | Joanna Kulig | film portrait (v4) |
jodiefoster |
woman | Jodie Foster | film portrait (v6), film portrait (v5), street portrait (v2), film scene (v2, 50) |
joelkinnaman |
man | Joel Kinnaman | film portrait (v5), film portrait (v4), street portrait (v2), film scene (v2, 50) |
joerogan |
man | Joe Rogan | - |
johncena |
man | John Cena | film portrait (v4) |
johnnydepp |
man | Johnny Depp | film portrait (v6), film portrait (v4) |
juliebenz |
woman | Julie Benz | - |
juliettelewis |
woman | Juliette Lewis | film portrait (v5), film portrait (v4) |
kaceyjordan |
woman | Kacey Jordan | - |
kaleycuoco |
woman | Kaley Cuoco | film portrait (v6), film portrait (v4), street portrait (v2), film scene (v2, 50) |
karenfukuhara |
woman | Karen Fukuhara | film portrait (v6), film portrait (v5), film portrait (v4) |
karengillan |
woman | Karen Gillan | film portrait (v5), film portrait (v4) |
karlurban |
man | Karl Urban | film portrait (v4) |
katdennings |
woman | Kat Dennings | film portrait (v5), film portrait (v4), street portrait (v2), film scene (v2, 50) |
katemara |
woman | Kate Mara | film portrait (v5), film portrait (v4) |
katemckinnon |
woman | Kate McKinnon | film portrait (v5), film portrait (v4) |
katewalsh |
woman | Kate Walsh | film portrait (v4) |
katewinslet |
woman | Kate Winslet | street portrait (v2), film scene (v2, 50) |
katieholmes |
woman | Katie Holmes | - |
keanureeves |
man | Keanu Reeves | street portrait, John Wick (50) |
keiraknightley |
woman | Keira Knightley | - |
kimdickens |
woman | Kim Dickens | - |
kirstendunst |
woman | Kirsten Dunst | film portrait (v5), film portrait (v4) |
kitharington |
man | Kit Harington | - |
klaudiajachira |
woman | Klaudia Jachira, speaker | film portrait (v4), parliament speech (50) |
kristenbell |
woman | Kristen Bell | - |
kristenstewart |
woman | Kristen Stewart | film portrait (v6), film portrait (v5), film portrait (v4) |
kristenwiig |
woman | Kristen Wiig | - |
krystenritter |
woman | Krysten Ritter | - |
kylieminogue |
woman | Kylie Minogue | - |
lainamorris |
woman | Laina Morris, YouTuber (Overly Attached Girlfriend) | film portrait (v4) |
leamichele |
woman | Lea Michele | film portrait (v4) |
leaseydoux |
woman | Lรฉa Seydoux | film portrait (v4) |
leonardodicaprio |
man | Leonardo DiCaprio | film portrait (v6), film portrait (v5), film portrait (v4) |
lindsaylohan |
woman | Lindsay Lohan | film portrait (v5), film portrait (v4) |
lisaedelstein |
woman | Lisa Edelstein | street portrait (v2), film scene (v2, 50) |
livtyler |
woman | Liv Tyler | - |
lucylawless |
woman | Lucy Lawless | - |
lucyliu |
woman | Lucy Liu | film portrait (v5), film portrait (v4) |
lupitanyongo |
woman | Lupita Nyong'o | film portrait (v6), film portrait (v5), film portrait (v4) |
madonna |
woman | Madonna | - |
maisiewilliams |
woman | Maisie Williams | film portrait (v4) |
maitlandward |
woman | Maitland Ward | - |
marieclaudebourbonnais |
woman | Marie-Claude Bourbonnais, model | - |
mattdamon |
man | Matt Damon | film portrait (v5) |
matthewmcconaughey |
man | Matthew McConaughey | film portrait (v6), film portrait (v5), film portrait (v4) |
mayahawke |
woman | Maya Hawke | film portrait (v5) |
melanielynskey |
woman | Melanie Lynskey | film portrait (v4) |
melgibson |
man | Mel Gibson | film portrait (v4) |
menasuvari |
woman | Mena Suvari | - |
miakhalifa |
woman | Mia Khalifa | - |
milakunis |
woman | Mila Kunis | film portrait (v5), film portrait (v4), street portrait (v2), film scene (v2, 50) |
mileycyrus |
woman | Miley Cyrus | Flowers video, Hannah Montana (50) |
millajovovich |
woman | Milla Jovovich | film portrait (v5) |
millyalcock |
woman | Milly Alcock | film portrait (v6), film portrait (v5) |
mireilleenos |
woman | Mireille Enos | film portrait (v6) |
morenabaccarin |
woman | Morena Baccarin | film portrait (v6), film portrait (v5), film portrait (v4), Deadpool, Firefly, Firefly (50) |
nathanfillion |
man | Nathan Fillion | street portrait, Castle, Firefly (50) |
nicolascage |
man | Nicolas Cage | film portrait (v6) |
noomirapace |
woman | Noomi Rapace | - |
normanreedus |
man | Norman Reedus | film portrait (v6), film portrait (v5) |
odetteannable |
woman | Odette Annable | - |
oliviawilde |
woman | Olivia Wilde | street portrait, House, Tron (50) |
orlabrady |
woman | Orla Brady | - |
pedropascal |
man | Pedro Pascal | film portrait (v6), film portrait (v5) |
rachelmcadams |
woman | Rachel McAdams | - |
rashidajones |
woman | Rashida Jones | film portrait (v6) |
rebeccaferguson |
woman | Rebecca Ferguson | film portrait (v6), film portrait (v5), film portrait (v4) |
rheaseehorn |
woman | Rhea Seehorn | film portrait (v6) |
rihanna |
woman | Rihanna | film portrait (v6), film portrait (v5), film portrait (v4), Ocean's 8 (50) |
robinwright |
woman | Robin Wright | - |
rosamundpike |
woman | Rosamund Pike | - |
rudaxsh |
woman | Rudaxsh, Instagram/TikTok | film portrait (v6), film portrait (v5), film portrait (v4) |
ryanreynolds |
man | Ryan Reynolds | film portrait (v6), film portrait (v5), film portrait (v4) |
sadiesink |
woman | Sadie Sink | film portrait (v6), film portrait (v5), film portrait (v4), street portrait, Stranger Things, Stranger Things (50) |
salmahayek |
woman | Salma Hayek | - |
samrockwell |
man | Sam Rockwell | film portrait (v6) |
samuelljackson |
man | Samuel L. Jackson | film portrait (v5), film portrait (v4), street portrait, Nick Fury, Pulp Fiction (50) |
sandrabullock |
woman | Sandra Bullock | - |
sarahhyland |
woman | Sarah Hyland | - |
sashagrey |
woman | Sasha Grey | film portrait (v5), street portrait, The Girlfriend Experience, The Girlfriend Experience (50) |
scarlettjohansson |
woman | Scarlett Johansson | film portrait (v4), street portrait, Black Widow, Black Widow (50) |
sigourneyweaver |
woman | Sigourney Weaver | film portrait (v5) |
sonoyamizuno |
woman | Sonoya Mizuno | - |
sophieturner |
woman | Sophie Turner | film portrait (v5), film portrait (v4) |
stevemcqueen |
man | Steve McQueen | - |
summerglau |
woman | Summer Glau | street portrait (v2), film scene (v2, 50) |
sweetanita |
woman | Sweet Anita, streamer | film portrait (v5), streaming (50) |
sydneysweeney |
woman | Sydney Sweeney | film portrait (v6), film portrait (v5), film portrait (v4) |
taylorswift |
woman | Taylor Swift | film portrait (v6), film portrait (v5), film portrait (v4), street portrait, Eras Tour close-up 1 (50), Eras Tour close-up 2 (50), Eras Tour close-up 3 (50), Eras Tour stage (50) |
terihatcher |
woman | Teri Hatcher | film portrait (v5) |
tildaswinton |
woman | Tilda Swinton | film portrait (v4) |
tomcruise |
man | Tom Cruise | film portrait (v5), film portrait (v4) |
tulsigabbard |
woman | Tulsi Gabbard | film portrait (v5), film portrait (v4) |
umathurman |
woman | Uma Thurman | - |
vindiesel |
man | Vin Diesel | - |
willaholland |
woman | Willa Holland | film portrait (v5), film portrait (v4) |
zendaya |
woman | Zendaya | film portrait (v5), film portrait (v4), street portrait, Dune (50) |
zoesaldana |
woman | Zoe Saldana | - |
Scenes without people, at 50 steps: alpine valley, macro dragonfly, Tokyo alley, rural village, vintage cars.
Two people in one shot
The training data had no group shots, yet two triggers in one prompt work well for a man and a woman: each keeps their own face. Name both with their class word, put them on a side, and describe each separately:
photo of nathanfillion, a man, and morenabaccarin, a woman, two people in the cockpit of a worn
frontier spaceship: on the left nathanfillion, a man with short dark hair and a wry smile, wearing a
brown leather duster coat over a red shirt; on the right morenabaccarin, a woman with long dark hair
pinned up in curls, wearing a flowing silk robe in deep jewel tones, ...
Samples (50 steps): Firefly, helicarrier, Dune, ship's bow 1, ship's bow 2, diner.
Two people of the same gender are hit-or-miss. Some work: the
stadium duet and the
palace audience keep two distinct women.
Others don't: in this library scene the
second woman has the right braids, collar and dress, yet a face pulled toward the first, and two men
on a beach do the same (the second loses most
of his likeness). Both triggers
share a class word (a woman or a man), which is likely where they meet. Costumes and hair always stay apart;
the faces are what can drift. Try several seeds, or render each person separately.
Three people in one shot
The same rule holds with three: faces stay distinct across class words, and drift within one.
- One woman, two men (a rainy city alley): all three keep their own faces.
- One man, two women (a spaceship cargo bay): the man is solid; the two women keep their costumes and hair, but their faces move partly toward each other.
- Three women (a castle dining hall): braids, red hair and house ties are all right, yet all three faces converge on the strongest identity of the three.
So for groups, mix genders where you can, and render same-gender groups with several seeds, or person by person.
The film scenes describe each role's look (costume, hair, setting) rather than naming the film, which the model does not know: see docs/generating.md.
The LoRAs
The lora-extract files hold the finetune as a LoRA for MOSS-Image Base: the
SVD of each layer's change, so they load on the base model and stack with other LoRAs. Use them
when you'd rather keep one base model and switch LoRAs. The workflow
lora-extract-turbo-lora.json is that setup: moss_image_base_bf16 โ serenity_v6_lora_r256
(1.0) โ the DMD turbo LoRA โ 12 steps, CFG 4, euler / simple.
They keep only part of each layer's change (by size, r64 drops 96%, r128 93%, r256 87%): at batch size 1 much of a full finetune's change is step-to-step noise that no low rank captures. That noise may not matter for likeness, but compare a LoRA with the full model on the same seed before relying on it. The full model keeps everything.
Training
Full finetune of MOSS-Image Base with maltrainer, summarised here and in full in docs/training.md:
| Trained | the Linear weights of the 42 joint blocks (4.76B of 5.08B parameters); the text adapter, norms and embedders frozen |
| Steps | 210,000 at batch size 1 (v6), in seven stages, 512 px aspect buckets |
| Optimizer | Automagic3, fused into backward, rate pinned at 1e-5 |
| EMA | 0.999 (updated every 4 steps, no warmup); the released weights are the average |
| Prior preservation | weight 1.0: each trigger is also trained against the original model's output for its class word, so "a woman" / "a man" stay as they were |
| Regularisation | real photos of women and men at 0.2ร the step |
| Timesteps | MOSS-Image's own final-pretraining distribution (logit-normal, mean 1.09, std 0.95, cutoff 0.95, 1% uniform) |
| Hardware | one RTX 5090 (32 GB): ~1.3 s/step, 24.9 GB peak |
Responsible use
This model can produce the likeness of real people. Do not use it to deceive, to impersonate, to harass, or to make sexual or demeaning images of anyone. Label generated images as generated.
License
Apache 2.0, as MOSS-Image. The DMD turbo LoRA used in the turbo workflows is lodestones' (lodestones/MOSS-Image-Base-8-steps-dmd) and is not part of this repository.
Model tree for malcolmrey/moss-serenity
Base model
OpenMOSS-Team/MOSS-Image-Base