Most AI porn workflows generate one action. You get a blowjob, or you get anal, or you get a cumshot. But real scenes have sequences — one act flows into the next. If you've been working with AI porn generation, you know that prompting multiple distinct actions in a single generation almost never works. The model picks one action and ignores the rest, or it blends them into an incoherent mess. The Ass to Mouth FFM workflow chains three completely separate generations to produce a full anal-to-deepthroat sequence from a single input image.

The fundamental problem is that each sexual action requires different body mechanics, camera framing, and motion patterns. Doggy anal involves rhythmic back-and-forth hip movement. A pull-out requires the camera to shift focus to the withdrawal. Deepthroat ass-to-mouth is an entirely different action involving a different performer with completely different head and throat motion. No single LoRA can handle all three, and stacking multiple LoRAs in one generation causes them to fight each other for control of what the model renders.

The Technical Breakdown

• 3-pass Image-to-Video workflow — each pass generates one distinct action with one dedicated LoRA
• Pass 1: fn_sm4shcut (0.75) + doggy_anal_cut (0.8) — smash cut into doggy anal position, 61 frames
• Pass 2: Body-Cumshot-Pullout (1.0) — pull out from ass, 61 frames
• Pass 3: Wan22_ThroatV3 (1.0) — deepthroat ass-to-mouth with the second woman, 101 frames
• Checkpoint: DasiwaWAN22I2V14B for all three passes
• All passes use WanImageToVideo — the last frame of each pass becomes the starting image for the next
• Post-processing: all three passes merged, upscaled 2x, RIFE interpolation 2x, rendered at 40fps
• Total output: approximately 11 seconds of seamless video

The innovation is the chaining itself. Three completely separate I2V generations run in sequence, and each one gets its own dedicated LoRA at the strength it needs to dominate the generation. The last frame of Pass 1 (anal) becomes the starting image for Pass 2 (pull out), and the last frame of Pass 2 becomes the starting image for Pass 3 (deepthroat). The result is three distinct actions that flow into each other as one continuous scene. No LoRA stacking, no prompt tricks — just clean handoffs between specialized models.

Workflow Specs

• Sampler: lcm
• Scheduler: beta
• Steps: 6
• CFG: 1
• Pass 1 frames: 61
• Pass 2 frames: 61
• Pass 3 frames: 101
• Pass 1 LoRAs: fn_sm4shcut (0.75) + doggy_anal_cut (0.8)
• Pass 2 LoRA: Body-Cumshot-Pullout (1.0)
• Pass 3 LoRA: Wan22_ThroatV3 (1.0)
• Upscale: 2x
• Frame interpolation: RIFE 2x
• Output FPS: 40
• Input: Upload an image of two women together

Note that Pass 1 is the only pass with two LoRAs. The smash cut LoRA (fn_sm4shcut at 0.75) handles the rapid transition into the position, while the doggy anal LoRA (doggy_anal_cut at 0.8) controls the actual penetration action. Both are kept below 1.0 so they cooperate rather than compete. Passes 2 and 3 each have a single LoRA at full strength for maximum control over their respective actions.

Related Workflows

• WAN 2.2 NSFW Workflows — Full Overview
• WAN 2.2 Blowjob & Deepthroat Workflow
• WAN 2.2 Smooth Sex Workflow
• WAN 2.2 Anal Insertion Workflow

How to Use This Workflow

1. Choose your input image carefully. Upload a photo of two women together. The first woman will be the one receiving anal in doggy position. The second woman will perform the ass-to-mouth deepthroat. Both should be clearly visible in the starting image.
2. Let the passes chain automatically. The workflow is pre-wired so that each pass feeds its last frame to the next pass as the starting image. You load one image and get the full three-act sequence without manual intervention.
3. Do not increase the Pass 1 LoRA strengths. The smash cut and doggy anal LoRAs in Pass 1 are deliberately set below 1.0 so they share control. Pushing either to 1.0 causes one action to overpower the other and the transition breaks.
4. The second woman should be visible in frame. Even though she is not active until Pass 3, the model needs to track her presence through Passes 1 and 2 so the transition to deepthroat feels natural rather than jarring.
5. Expect variation between runs. Three chained passes mean three points where generation randomness compounds. Some runs produce perfect seamless transitions, others may have slight discontinuities at the handoff points. Generate a few runs and keep the best one.

Want to create your own ass-to-mouth FFM scenes? Run this workflow in the VirtuaVixen AI Studio with your own images, or download the full workflow file from the VirtuaVixen Pro ComfyUI Workflow Pack to run it on your local ComfyUI setup.