Image generation models
The three text-to-image models on Sawala Cloud — who makes them, what drives their cost, and which one to start with.
Image generation turns a written description into a picture. On Sawala Cloud this happens inside a flow: an image step takes a prompt — assembled from your own content if you like — and returns an image the flow can attach to a message, store as a file, or publish.
Two things drive what an image costs. The first is its size: bigger images cost proportionally more. The second is the number of refinement passes (also called diffusion steps) the model makes — each pass sharpens the result, and each one is billed. A model designed for few passes is therefore dramatically cheaper than one that needs many.
FLUX.1 Schnell
Made by Black Forest Labs, the lab founded by the original Stable Diffusion researchers. Schnell is German for "fast", and that is the point of this model.
It is a turbo model: engineered to produce a finished image in very few passes, and it accepts at most 8. That makes it both the fastest and, by a wide margin, the cheapest image model on the platform. It emits a fixed image size — you cannot ask it for a specific width, height, or aspect ratio.
This is the default, and the right first choice. Use it for anything where throughput and cost matter more than precise dimensions.
Available in Flow only — chat surfaces never transcribe, draw, or speak, so this model does not appear in the Crew or Connect model pickers.
Leonardo Lucid Origin
Made by Leonardo.Ai, an Australian studio whose models are aimed at design and marketing work.
Its reason to exist here is custom output size: unlike FLUX, it accepts a width and height, so you can ask for a landscape banner, a square post, or a portrait card. It is not a turbo model — it needs roughly 20 to 40 passes to produce a sharp, finished image, and at low pass counts the result looks soft and unfinished.
That combination makes it many times more expensive per image than FLUX.1 Schnell. Choose it when you need a specific aspect ratio or higher fidelity, and expect to pay for it.
Available in Flow only — the same reason as above.
Leonardo Phoenix 1.0
Made by Leonardo.Ai, from the same family as Lucid Origin and with the same shape: custom width and height, roughly 20 to 40 passes for a finished result, and a comparable cost per image.
Treat it as a design-oriented alternative rather than an upgrade — it is a different aesthetic, not a different capability. If Lucid Origin is not giving you the look you want, try this one with the same prompt.
Available in Flow only — the same reason as above.
Start with FLUX.1 Schnell. It is the default, the fastest, and the cheapest, and its fixed output size is fine for most uses. Move to a Leonardo model only when you need a particular aspect ratio or a more polished result — and when you do, remember to raise the pass count, or the image will look unfinished.
Next: speech synthesis models.
Model list last verified: 2026-08-07.