When nobody has used the GPU for a while, the server is cold. The first job you start needs to boot the hardware before it can actually render.

Each “standard” clip in this studio is around 9 seconds of video, but the GPU needs more time to process it:

• Warm server: ~308 seconds per clip (about 5.1 minutes).
• Cold server (first clip): ~800 seconds total (about 13.3 minutes) for warm-up + render.

Pricing is flat per render: each successful generation costs 40 tokens, whether the server is cold or warm. Cold vs warm only affects how long you wait, not how many tokens are used.

What you will notice

• The job status may stay on “starting” / “preparing” longer on the first run.
• The first run after a longer break takes much longer than the next ones.

Once you click the button to start a workflow, the job is sent to the server and begins processing. At this point:

Do not:

• Do not refresh the page.
• Do not close the browser tab.
• Do not open the same workflow in the same tab and click run again.

Refreshing or closing the tab while a job is running can interrupt the connection to the server and may cause:

• Lost job status or missing previews.
• Failed renders or duplicate runs.
• Tokens being consumed without a clean result in the UI.

If you want to process more renders, use the buttons in the interface that are meant for starting another job. Do not spam refresh or re-run the same job while it’s still active.

Recommended workflow

• Start one job and wait until it finishes.
• Use the “Run Again”, “New Job”, or “Queue Another Render” button (depending on how it’s labeled) to launch the next one.
• Only change settings between jobs, not during a running job.

The Awesome Fucking and Awesome Sucking workflows are built as flexible “all-rounders” for explicit scenes.

What they are good at

• Handle a wide range of body positions and camera angles.
• Optimized prompts and settings are already pre-loaded.
• No need to fight the model with extreme or overcomplicated wording.

How to prompt them

• Describe the vibe, style, camera angle, and setting.
• You can let the default prompt handle the “NSFW part”.
• Add only the key details you really care about (lighting, outfits, style, etc.).

A good prompt for this studio is clear and focused, not a huge wall of text. The internal setup already does a lot of heavy lifting.

Keep it structured

• Who: model look (hair, body type, outfit, energy).
• Where: environment / room / background.
• How: mood, lighting, style (cinematic, glossy, anime, etc.).
• Camera: close-up, mid-shot, wide, from above, from below, etc.

Things that help

Things that hurt

To keep things simple, the studio uses a flat price per render. Every successful generation of a standard ~9 second clip costs:

40 tokens per render

This price is the same whether the server is cold or warm. Cold starts only affect how long you wait, not how many tokens are deducted.

Why batching still helps

If you start from a cold server and run several clips in a row with the same workflow, you pay 40 tokens for each render either way — but only the first one has the long warm-up. The next clips use the already-warm GPU and finish much faster.

• 1 clip after a long break → 40 tokens, longest wait.
• 3–5 clips back to back → 40 tokens each, only the first one is slow.
• Long batches (many clips) → best use of your time, same flat cost per render.

How many renders per pack?*

• 50 tokens ≈ 1 render (10 tokens left).
• 250 tokens ≈ 6 renders (10 tokens left).
• 900 tokens ≈ 22 renders (20 tokens left).
• 1,800 tokens ≈ 45 renders.
• 3,000 tokens ≈ 75 renders.
• 10,000 tokens ≈ 250 renders.

*Calculations assume 40 tokens per generation. Future non-standard workflows (different durations or tools) may use different pricing, but the studio will always show how many tokens are required before you run.

Browser & device

• Use a modern desktop browser (Chrome, Edge, Firefox, or similar).
• Avoid heavy multitasking in the same browser while a job is running.
• Disable aggressive ad-blockers for this page if the UI doesn’t respond correctly.

Resolution & performance

• Higher resolutions look better but take longer and may fail more easily.
• If a job keeps failing, try a slightly lower resolution or shorter duration (for video workflows).

Tokens & safety

• Every completed run consumes 40 tokens (unless otherwise stated in the UI).
• Spam-clicking the run button can start multiple jobs and burn tokens quickly.
• Always wait for a clear success or error before starting another job.

Job feels stuck on cold start

• Give it a few minutes, especially if it’s the first job in a while.
• Check that your internet connection is stable.

Output looks wrong or off-style

• Reset to the default prompt and run once.
• Only change 1–2 things at a time (e.g. camera angle + outfit).
• Avoid mixing too many conflicting style tags.

You lost the page

• If you accidentally refreshed or closed the tab, wait a bit and then try again with a new job.
• To avoid this next time, keep the job tab open until you have your final result saved.