Vivo X80 Pro “RAW” looks cooked - anyone here actually get a truly single-frame DNG?
After a year with the X80 Pro, I’m noticing artifacts that don’t line up with a clean sensor dump. DNGs from the stock app (and some third‑party Camera2 apps) show multi-frame ghosting, field-of-view shifts, and lens shading that looks pre-corrected. If the V1+ is touching the pipeline before the DNG is written, that undermines the whole point of shooting RAW for a lot of workflows.
What I’m seeing:
- Motion ghosting in DNGs that doesn’t match the EXIF shutter (e.g., ceiling fan blades “double” even at 1/250s).
- Slight FOV mismatch between DNG and JPEG as if EIS/cropping is applied to one and not the other.
- Periscope DNGs with vignetting oddly tamed, suggesting lens shading correction baked in.
- Noise texture that looks spatially correlated like temporal NR, not sensor grain.
- Black level/white level metadata inconsistent with true RAW, plus odd ColorMatrix entries vs the sensor’s expected CFA.
Questions for X80 Pro owners and camera devs:
- Is vivo doing multi-frame fusion and/or NR on DNG even with all “enhancements” toggled off in Pro mode?
- Does the V1+ ISP engage for RAW capture, or only for preview/JPEG? Any known method to force a pass-through RAW stream?
- Why do third‑party apps via Camera2 produce flatter DR but worse motion than stock? Are vendor tags (e.g., multiframe/NR flags) exposed anywhere?
- Has anything changed on recent FuntouchOS updates (13 → 14 → 14.x)? Any build where RAW became “purer”?
- Is there a way to access RAW10/RAW12 streams that bypass the computational path (e.g., via GCam mods or hidden Camera2 combinations)?
- On the periscope, are we seeing OIS pose data being used to align frames pre-DNG? That would explain the ghosting with fine detail.
Simple, reproducible tests we can crowd-run:
- LED flicker test: shoot DNG under 50/60 Hz lighting at 1/100s, 1/50s, 1/25s. True single-frame RAW should show predictable banding per shutter; multi-frame RAW often shows smeared/uneven bands or inconsistent EXIF vs visual result.
- Fan blade test: high-speed fan at 1/500s RAW vs JPEG. Look for double edges in DNG that indicate temporal stacking.
- Star field test: point at the night sky at 1-2s exposures. Check for micro‑trails/ghost stars unique to DNG.
- Lens shading check: shoot a uniformly lit surface. RAW should show clear vignetting; if it’s flat, shading is likely pre-applied.
- Metadata dive: dump CameraCharacteristics/EXIF (Camera2 Probe, exiftool). Check RAW capability level (FULL/LEVEL_3), blackLevelPattern, noiseProfile, opticalBlackRegions, pipelineDepth, and any vendor tags. Compare across main/ultra‑wide/periscope.
What to share if you can help:
- Three matched pairs (JPEG + DNG) from stock camera Pro mode with all enhancements off, across main and periscope, including one motion scene and one uniform surface.
- Same scenes from a third‑party app (Open Camera, ProShot, GCam mod) with RAW enabled.
- Build number/FuntouchOS version and camera app version.
If these DNGs are indeed computational composites, should OEMs label them “Computational RAW” and give a toggle for true single-frame RAW? I’m not against multi-frame for average users, but hiding it behind a “RAW” label breaks pro workflows. If someone has a known combo of settings/ports that yields genuine single-frame RAW on the X80 Pro, post it.