🎬 Object Focus Camera v7 (Pro Cinema)

Object Focus Camera V7 ("Pro Cinema") is the version for people who think in shot lists. It swaps V5's "close/medium/far" distance labels for actual cinematography vocabulary - Extreme Close-Up through Extreme Wide Shot, pan/tilt/dolly/truck/pedestal, 14-24mm ultra-wide through anamorphic - because Qwen Edit responds to named camera concepts better than to vague distance words. If you've ever wanted a spec-sheet camera instruction without going to film school, this is the node.

The panel, decoded

The heart of V7 is shot_size, which replaces the distance dropdown with the eight standard classifications - ECU, CU, MCU, MS, MLS, FS, WS, EWS. The tooltip maps each one to a real distance (ECU ≈ 0.3m, CU ≈ 0.8m, MS ≈ 2.0m, EWS ≈ 10.0m), and the node turns that into words the model understands. That mapping is the clever bit: film terminology is just distance with intent, and this node does the conversion for you.

Around that:

• camera_angle (35 choices) - eye level through worm's-eye, dutch angle, plus every angled view from 15° to 60°.
• camera_movement (20 choices) - the complete professional set: pans, tilts, dollies, trucks, pedestals, arcs, plus the "---" category headers so you can find them.
• height and direction (8 each) - used when you pick a custom "Vantage" position: ground_level to bird_eye, front to back and diagonals.
• framing_mode - Shot Size Presets or Custom Meters. Pick Custom and distance_meters (0.3–15m) takes over, auto-converted to words.
• auto_facing - keeps the camera aimed at the target for custom vantage positions. Leave it on.
• lens_type (12) - 50mm normal, 85mm portrait, 100-200mm telephoto, macro, fisheye, tilt-shift, anamorphic, soft focus. The anamorphic option is a fun one to try for that letterboxed cinematic stretch look.
• prompt_language - Chinese/Hybrid if you run dx8152's Qwen Edit LoRAs (they're Mandarin-trained), English/Universal otherwise.
• focus_transition_mode - defaults to Focus Transition here, i.e. V7 assumes you want the camera to reposition onto the object rather than hold its spot.
• add_detailed_explanation - defaults to Detailed (Full cinematography explanation), the most verbose setting, because this node is about articulating camera intent.
• material_detail_preset (37 options) and photography_quality_preset (15 options) - inherited from V5, still there for the finish.

How it works

Plain-language Qwen prompt composition, but layered deeper than any sibling: shot size → angle → movement → vantage height/direction → lens → material → quality → explanation. The prompt output is ready for the encoder; system_prompt is the preserve-the-scene camera-operator framing (in transition mode especially, the scene has to survive a camera journey); description is your shot-list summary.

The honest take

V7 is overkill for a quick product shot - V3 or V5 do that in a fraction of the dropdowns. It earns its keep on the workflow where you're generating a sequence: a walkthrough, a flyaround, a series of shots at different sizes and angles that you feed frame by frame into Qwen Edit. That's exactly the multi-frame use case the README points at with its FLF video roadmap. Just remember the family physics: exact angles are approximate, direction is reliable, and the model can't actually do lens compression - "anamorphic" is a look, not optics.

Install

Same pack, same drill - ComfyUI Manager ("ArchAi3d Qwen") or:

cd ComfyUI/custom_nodes && git clone https://github.com/amir84ferdos/ComfyUI-ArchAi3d-Qwen.git
cd ComfyUI-ArchAi3d-Qwen && pip install -r requirements.txt

Restart, find it under ArchAi3d/Qwen/Camera. You need the Qwen-Image-Edit weights (quantized on consumer GPUs). And the license note applies here louder than anywhere: free for personal work, paid client projects need the commercial license.