Blog/prompts·21. Nov. 2026·8 Min.·vom eroq-Team
Eroq One prompt guide — ten reference pictures and edits you write in a sentence
How to prompt Eroq One on eroq: give each reference a job, write edits as one change, combine faces, outfits and products, with 11 examples and the price per picture.
Eroq One is eroq's reference engine: it builds one picture from up to ten references (faces, outfits, a product, a place) and edits an existing picture from a plain-language instruction. The rule that matters most is to give every picture a job in the prompt: "the face from @Picture 1, the jacket from @Picture 2, the street from @Picture 3".
That is what separates it from the rest of the image lineup. Krea2 and Eroq Uncensored read up to four pictures and re-render around them; Eroq One reads ten and composes from all of them, and its editor changes what you name while the untouched pixels stay where they were. It is uncensored within the acceptable-use policy and runs fully private. This guide covers both ways of prompting it, the verbs the editor understands, eleven examples, the price per picture and the mistakes that waste a take.
What Eroq One is good at, and what it is not
- Consistency across a series: the same character, outfit and product from image to image, because they ride in as pictures rather than as descriptions.
- Composition from parts: a person from one photo, clothes from another, a product from a pack shot, a place from a location photo, one picture out.
- Instruction edits: "make it night", "swap the jacket for red leather", "remove the man in the background", with the rest of the picture kept.
- Frames for film: Cinema uses it to paint a frame with your cast's looks.
It is slower than the single-reference engines when you load it up (about ten seconds from text, twenty to fifty with pictures), it does not set lettering, and it is not the anime engine. For fast portrait iteration with one face, Krea2 is quicker; for words on a picture, use Flash Image.
Two modes, two kinds of prompt
The Image tool has a Create tab and an Edit tab, and Eroq One wants a different sentence in each.
Create: describe the shot, assign the pictures
Drop pictures in (they number themselves Picture 1, Picture 2…), or @mention saved Characters and Elements, whose pictures ride along. Then write the scene and say which picture supplies what:
[shot and action] — [who] from @Picture 1, wearing [the outfit in @Picture 2], with [the object in @Picture 3], in [the place in @Picture 4], [light], [camera]
The order of the pictures matters less than the sentence that points at them. A reference that nobody mentions is a reference the engine has to guess about.
Edit: write the change, not the scene
In the Edit tab, Picture 1 is the one that changes; any other pictures guide the change. The prompt is an instruction, ideally one sentence:
[verb] [what] [to what], keep [what must not move]
You do not re-describe the room, the person or the light. You name the change and, when it matters, what has to stay. The edit keeps your picture's frame: Format shows Retain, and the result comes back at the engine's usual resolution in the source's ratio.
The verbs the editor reads well
| Verb | Example |
|---|---|
| Change / swap | "Change her jacket to red leather, keep everything else" |
| Add | "Add light rain and wet reflections on the pavement" |
| Remove | "Remove the passer-by on the right and fill in the shop window behind him" |
| Make it | "Make it night, with the street lamps on and the windows lit" |
| Turn / move | "Turn his head toward the camera, same expression" |
| Replace with | "Replace the label on the bottle with the one in @Picture 2" |
| Restyle | "Restyle as a 1970s film photograph, warm and grainy, same framing" |
One verb per pass works best. Chain edits instead of stacking them: night first, then rain, then the jacket. Each pass is a take you can keep or discard.
The dial
Prompt strength (CFG, 0 to 3, Loose to Strict, default 1) sets how hard your words push. In Edit, raise it a step when the change is too timid; lower it when the edit spills into areas you did not mention. There is no negative prompt on Eroq One: describe the result you want.
Eleven example prompts
Created from references
1. Character, outfit, place — three pictures, three jobs.
Full-length shot of the woman from @Picture 1 walking toward camera on a wet cobbled street, wearing the long camel coat from @Picture 2, in the old town square from @Picture 3, overcast daylight after rain, 35mm, natural color
Settings: 3:4, four takes, three pictures.
2. Product in hand — the label survives because it rides as a picture.
Close-up of the hand from @Picture 1 holding the amber glass bottle from @Picture 2 at eye level, label facing camera, blurred bar shelves behind, warm practical light, 85mm, shallow depth of field
Settings: 1:1, four takes, two pictures.
3. Two Characters and a set — saved Characters plus a location Element.
@Mara and @Theo arguing across the counter of the diner from @Picture 1, Mara slamming down a coffee pot, Theo leaning back with his arms crossed, late-night fluorescent light, wide angle, film grain
Settings: 16:9, two takes, two Characters and one place.
4. A look sheet in a new pose — the sheet carries identity; the prompt carries the moment.
@Ines sitting on the bonnet of a vintage car at a desert gas station at golden hour, sunglasses pushed up into her hair, squinting into the sun, dust in the air, 50mm, warm film look
Settings: 16:9, four takes, one Character.
5. Mood board to scene — a style picture sets color and texture, not content.
A fisherman in his sixties mending nets on a harbor wall at dawn, in the color palette and grain of @Picture 2, his face from @Picture 1, gulls overhead, soft side light, 35mm
Settings: 4:3, two takes, two pictures.
Edits
6. Recolor, keep everything else — the classic one-line edit.
Change his denim jacket to red leather, keep his face, pose, the street and the light exactly as they are
Settings: Edit, Retain, one take, Picture 1 only.
7. Day to night — name the light sources that should switch on.
Make it night: street lamps on, warm light in the café windows, a deep blue sky, the people and their positions unchanged
Settings: Edit, Retain, two takes.
8. Clean the background — say what fills the gap.
Remove the man walking behind her and continue the brick wall and the bicycle rack where he was
Settings: Edit, Retain, two takes.
9. Weather — a small edit that changes the whole mood.
Add steady rain, wet hair and darker wet pavement with reflections, keep the framing and her expression
Settings: Edit, Retain, two takes, Prompt strength 1.2.
10. Expression — faces are where an editor proves itself.
Change her expression to a quiet, closed-mouth smile, eyes still on the camera, nothing else changes
Settings: Edit, Retain, four takes.
11. Swap a label from another picture — the second picture guides the first.
Replace the label on the jar with the label in @Picture 2, same curve and lighting on the glass
Settings: Edit, Retain, two takes, two pictures.
How pictures are priced
Eroq One is priced by the pictures it reads: 10 credits / image, +1 per extra picture. From text or a single picture, a take is 10 credits (an edit counts its source picture), each extra picture adds 1 credit, so four pictures cost 13 credits and the full ten cost 19. Each take is priced the same way, and the Generate button shows the total before you click. A failed render refunds itself.
In practice: load the pictures that carry identity (faces, a product, a signature outfit) and describe the rest. A location you can describe in eight words does not need to be a picture.
Mistakes to avoid
- Unassigned pictures. Five references and a prompt that mentions none of them leaves the engine to decide who is who. Point at each one.
- Re-describing the scene in an edit. "A woman on a street at night in a red jacket…" turns an edit into a fresh render. Write only the change.
- Several edits at once. "Make it night, add rain, change the coat and remove the dog" in one pass loses some of the four. Chain them.
- Conflicting references. Two photos of the same person with different hair colors make a blend. Choose the one you want.
- Expecting a new format in Edit. The editor keeps your picture's frame. To change the ratio, generate in Create with the picture as a reference instead.
- Lettering. Signs and labels written from scratch come out garbled; swap in a real label from a picture instead (example 11), or use an engine that sets text.
Settings on eroq
- Formats: 1:1, 3:4, 4:3, 9:16 and 16:9 for new images; Retain in Edit.
- Takes: one to four per run.
- References: up to ten pictures, or your Characters and Elements.
- Access: every account, pay-as-you-go included.
- Content: uncensored within the acceptable-use policy: fiction only, nothing involving minors, no real identifiable people without consent.
- Privacy: eroq runs the engine itself and never trains on your prompts or pictures. The Private mode switch at the bottom of Settings keeps the result in your browser only; reference pictures added in Private mode never leave the tab.
FAQ
Is Eroq One free to try?
Yes: a new account gets 40 free credits with no card. A take from text or one picture costs 10 credits, so that is a handful of tries before you pick a pack (from $10) or a plan.
How many reference pictures can Eroq One read?
Up to ten, from uploads, your Library, or your saved Characters and Elements. Each one can carry a face, an outfit, a product or a place.
Why did my edit change something I did not ask for?
The instruction was probably broad, or Prompt strength was high. Name what must stay ("keep his face and the background"), make one change per pass, and lower Prompt strength a step if the edit still spills.
Can Eroq One keep my picture's aspect ratio?
In Edit, yes: it always keeps the source's frame (Format shows Retain). In Create, pick any of the five formats and your pictures ride in as references.
When should I use Eroq One instead of Krea2?
When more than four pictures matter, when an edit must leave the rest of the picture untouched, or when a product or outfit has to come back identical. For quick single-face portraits, Krea2 is faster at the same base price.
Open the Image tool and pick Eroq One: /studio/image.

Render es in Bild
Alles oben passiert im Studio: dieselben Engines, dieselben Einstellungen, gerendert auf deinen Klick.
40 Gratis-Credits · keine Kreditkarte · Privatmodus für jedes Konto
Prompts aus diesem Artikel
[shot and action] — [who] from @Picture 1, wearing [the outfit in @Picture 2], with [the object in @Picture 3], in [the place in @Picture 4], [light], [camera]
Privatsphäre
Privat und uneingeschränkt
- →Kein Filter auf den eigenen Engines von eroq: stattdessen eine schriftliche Richtlinie.
- →Sie behalten nichts, und nichts wird mit deiner Arbeit trainiert.
- →Der Privatmodus speichert keine einzige Datei.