Promote rectangular latent upscale baseline
This commit is contained in:
@@ -62,6 +62,7 @@ The initial aspect ratio must remain 16:9 and the same final target must be used
|
||||
- SLU-15: best upscale ceiling + best tile + best initial dimensions
|
||||
- SLU-16: quality-first combination, accepting a longer runtime if the visual improvement is clear
|
||||
- SLU-17: practical production combination targeting a predictable sub-hour render
|
||||
- SLU-18: rectangular 1024×640 tile shape with four effective tile passes
|
||||
|
||||
## First two runs
|
||||
|
||||
@@ -84,11 +85,12 @@ These are the valid runs recorded so far. Runtime is wall-clock generation time;
|
||||
| SLU-07 | 672×384 → 1920×1088 | 768×768 | 26m 01s | **12.45 min/Mpx** | 1920×1088 | **Current best** |
|
||||
| SLU-08 | 672×384 → 1920×1088 | 864×480 | 31m 13s | 14.93 min/Mpx | 1920×1088 | Valid; visual comparison pending |
|
||||
| SLU-11 | 768×432 → 1920×1088 | 768×768 | 31m 07s | 14.91 min/Mpx | 1920×1088 | Valid; review pending |
|
||||
| SLU-18 | 672×384 → 1920×1088 | 1024×640 | **25m 14s** | **12.08 min/Mpx** | 1920×1088 | **Reviewed; new current best** |
|
||||
|
||||
### Current best
|
||||
|
||||
**SLU-07 is now the current best baseline**: 672×384 initial generation, 768×768 tiles, 128-pixel overlap and a 1920×1088 output. It was judged good quality, better than the original 1080p baseline, with stable motion and no objectionable seams. At 26m 01s it is also the fastest valid 1920×1088 result and has the best normalized throughput so far. SLU-06 remains a good fallback at 26m 52s. SLU-11 did not improve runtime because its larger initial generation took 31m 07s.
|
||||
**SLU-18 is now the current best baseline**: 672×384 initial generation, 1024×640 tiles in a 2×2 layout, 128-pixel overlap and a 1920×1088 output. It was reviewed and accepted as good quality. At 25m 14s it is 47 seconds faster than SLU-07 and has the best normalized throughput so far at 12.08 min/Mpx. The attention `keep_percent 5` test was slower at 31m 05s, so the baseline value of 10 remains in use. SLU-06 remains a good fallback at 26m 52s. SLU-11 did not improve runtime because its larger initial generation took 31m 07s.
|
||||
|
||||
SLU-09 full-frame/no-split failed with GPU out-of-memory on the 8 GB card and is not a viable baseline. The tracked `workflow.json` now uses the SLU-07 settings. Static prompt, seed, references, model, samplers, schedulers, steps, denoise and final target remain unchanged.
|
||||
SLU-09 full-frame/no-split failed with GPU out-of-memory on the 8 GB card and is not a viable baseline. The tracked `workflow.json` now uses the SLU-18 settings. Static prompt, seed, references, model, samplers, schedulers, steps, denoise and final target remain unchanged.
|
||||
|
||||
The earlier tile submissions that produced 1024×1024 output are excluded from this table because their target-resolution override was not applied.
|
||||
|
||||
Reference in New Issue
Block a user