by AlibabaImage
Z-Image Base LoRA
Z-Image Base with up to three of your own LoRAs
- Up to 1.5K
- 9 aspect ratios
- Loads your own LoRA
From 3 credits per image≈ $0.03· Create Image
What is Z-Image Base LoRA?
Z-Image Base LoRA runs the full 6 billion parameter Z-Image base model from Alibaba's Tongyi team with up to three of your own LoRAs. Base is the build recommended for training, so LoRAs trained on it load here at full quality, for text to image in nine shapes and image to image with a strength control.
Best on PopNow for: Testing LoRAs you trained on Z-Image Base, turning a set of photos into matching illustrations, and photographic portraits and cinematic 21:9 frames.
Good for
- Testing LoRAs you trained on Z-Image Base
- Turning a set of photos into matching illustrations
- Photographic portraits and cinematic 21:9 frames
What Z-Image Base LoRA does well
Built for your trained LoRAs
If you trained on Z-Image Base, this is where your LoRA performs as intended, at full quality.
Photographic without any LoRA
On its own the base model draws natural light, skin and grain from a few sentences.
Photo to illustration
At strength around 0.6, an illustration LoRA redraws a photo while keeping the pose and setting.
Real results from Z-Image Base LoRA
Every image and clip here was generated with Z-Image Base LoRA. Hit Recreate to open it in the studio with the same prompt and settings.
InputBring your own image
InputBring your own image
Prompts that worked on Z-Image Base LoRA
Real outputs with the exact prompt we used, and why it worked. Hit Recreate to try the same settings.
mature illustration with muted color and complex linework. An East Asian woman in her thirties with a long black ponytail, in a sleeveless black top and a leather arm guard, draws a recurve bow in a pine forest clearing at sunrise, gold shafts of light through the trunks and mist, ferns at her feet.
The LoRA's style phrase (mature illustration with muted colour and complex linework) opens the prompt, followed by a faithful description of the photo. That pairing keeps the composition while changing the medium.
RecreateA photograph of a woman in her early thirties with dark brown skin and short natural curls, wearing fitted cream riding breeches, tall black boots and a loose white linen shirt with rolled sleeves and an open collar, leaning against a wooden stall door in a stable at first light while a chestnut horse rests its head beside her shoulder. Low sun cuts through the dusty air in long golden shafts, straw on the floor, warm film grain, 35mm lens, shallow depth of field.
No LoRA: a plain photographic description of person, clothing and light. The base model's own photo look is strong enough for portraits.
RecreateWhat Z-Image Base LoRA can do
Each mode opens in the studio with Z-Image Base LoRA already picked. The price shows before you generate.
Best ways to use Z-Image Base LoRA on PopNow
Where this model already does the work for you, from one-tap apps to the full studio.
How to prompt Z-Image Base LoRA
- Use LoRAs trained on Z-Image Base. Turbo LoRAs can break the image here without an error.
- Run the same prompt and seed with and without the LoRA to check it's adding something.
- For photo to illustration, use a strength around 0.6; from about 0.75 the pose and setting get redrawn.
Settings worth knowing
Everything else has a sensible default, so you can type a prompt and go.
resolution1K · 1.5K- Output size. Bigger sizes look sharper and cost more.
aspect_ratio1:1 · 16:9 · 9:16 · 4:3 +5- The shape of the frame: wide, square or tall.
image_url- The picture to start from.
seed0–2147483647- Reuse the same number to get a repeatable result.
loras≤ 3- Load your own LoRA files to lock in a style or character.
strength0–1- How far the result may move away from your input.
output_formatjpeg · png · webp- File type of the result.
How it compares
Same kind of job, different strengths and prices.
Z-Image Turbo LoRA is much faster and suits LoRAs trained for the Turbo build.
Where Z-Image Base LoRA falls short
There's no negative prompt, so describe what you want rather than what to avoid.
- Turbo LoRAs can break the image here without an error.
- There is no negative prompt.
- Image to image redraws the whole frame; there is no mask.
How to use Z-Image Base LoRA in three steps
- Pick Z-Image Base LoRA in the image studio and add up to three Base-trained LoRA links.
- Describe what you want, since there is no negative prompt.
- For photo to illustration, upload the photo and set strength near 0.6.
Questions about Z-Image Base LoRA
How much does Z-Image Base LoRA cost?
On PopNow, the cheapest run of Z-Image Base LoRA costs 3 credits ($0.03) per image. You pay with credits (1 credit = $0.01), the exact cost shows on the Generate button before you start, and failed generations are refunded automatically.
What sizes can Z-Image Base LoRA make?
Z-Image Base LoRA renders up to 1.5K in 9 aspect ratios.
How is Z-Image Base different from Z-Image Turbo?
Base is the full, undistilled model recommended for training; Turbo is much faster.
Do Z-Image Turbo LoRAs work on Base?
Not reliably; in tests a Turbo LoRA broke the image without an error.
What strength should I use for photo to illustration?
Around 0.6. From about 0.75 the pose and setting get redrawn.
How do I check a LoRA is working?
Run the same prompt and seed with and without it and compare.
What does it cost?
A low flat price per image for both modes.
Z-Image Base LoRA or Z-Image Turbo LoRA?
Z-Image Turbo LoRA is much faster and suits LoRAs trained for the Turbo build.
Make something with Z-Image Base LoRA
Trained a LoRA on Z-Image Base? This is where it looks its best.