Every Dial in the Studio, and Which Ones Actually Change the Result

Every generator ships a row of controls and almost none of them explain what the control actually does, so people develop habits: always High quality, always PNG, always 4K, always the biggest number available. That gets expensive fast and improves surprisingly little.

Here is the whole surface, what each control genuinely changes, and an honest ranking at the end. One thing to say up front, because the phrase gets used loosely in this industry: there is no model fine-tuning here. No LoRAs, no training on your own images, no custom checkpoints. What you have is control over each generation, which is a different and more immediate thing.

Images: three controls

Quality. Standard renders at 1.5K for 40 coins, High at 2K for 60. The difference shows in fine texture, skin and fabric mostly, and not at all in composition. A badly composed picture at 2K is a bigger badly composed picture. Explore at Standard, finish at High. The same logic in more detail is in Standard or High, PNG or JPEG.

Format. PNG is lossless and large; JPEG is compressed and small. If the image is going into an editor for heavy work, take PNG. If you are looking at it, sending it, or using it as a video start frame, JPEG is visually identical and a fraction of the size.

Aspect ratio. Sixteen options plus auto. This one has more effect on the picture than either of the others, because the model composes for the shape it is given. Ask for a portrait in 16:9 and you get a person marooned in scenery; ask in 4:5 and you get a portrait. If you do not have a strong opinion, auto gives the model a sensible choice.

Video: six controls

Model. The biggest lever by far, and the one most people never touch. There are four: image-to-video and text-to-video, each in a fast turbo variant and a slower, sharper full one. Text-to-video invents the whole scene from a sentence; image-to-video takes a still you have already approved and moves it. For a specific person the image path wins almost always, because the face is already settled.

Resolution. Priced per second, and the spread is enormous: turbo charges 125 coins per second at 720p and 150 at 1080p, while the full model runs 115 at 480p, 225 at 720p, 550 at 1080p and 1100 at 4K. Treat 4K as a finishing pass on a shot you have already proven.

Duration. Anywhere from 4 to 30 seconds, charged by the second. Longer is not safer: the model has more time to drift, so a long clip is a bigger bet, not a better one.

Aspect ratio. Seven options plus auto. On video, auto means keep the start image's framing, which is the right default far more often than a chosen ratio.

Audio. On or off, free either way. Off is right whenever you intend to add your own sound later.

Start and end frame. The one almost nobody uses. Give a clip both a first and a last image and the model builds the motion between them instead of improvising where the movement lands. It is the difference between describing a movement and directing one. Full details are in what Fellowi Video can actually do.

The control that is not a setting

A reference image does more for consistency than every dial combined. Text cannot carry a face: two generations from one description give you two different people, no matter how many adjectives you add. Hand the generator a picture instead and it keeps the likeness. That is how a companion's photos stay recognisably her across fifty sends, and it is available to you directly through the video start frame. The method is in keeping a character consistent.

The honest ranking

1. The prompt. Nothing else is close. A specific scene with a place, a light source and a pose beats every setting change you could make. Start with how to write prompts that work, or the explicit version in writing NSFW prompts that render.
2. A reference image, if you care who is in the picture.
3. Aspect ratio, because it changes composition rather than just size.
4. Model, on video. Turbo versus full is a real quality difference at a real price difference.
5. Everything else: quality tier, format, duration, audio. These decide what you get and what it costs. They do not decide whether it is good.

And the setting that makes all of this affordable is not a setting at all: a failed generation is refunded in full, automatically, so the correct way to use the controls is to try the cheap combination first and pay for the expensive one only once you know the shot works. Open the studio and change one dial at a time.