OD SETTINGS AND TEST LOG

Status: Not run / Blocked / Inconclusive / Observed pass / Observed failure
Test ID and attempt number:
Date and time (include time zone):
Character and revision:
New practice chat or existing accepted timeline:
Chat or media route:
Actual selected model label:
Device/browser and visible account controls, when relevant:
Source guide URL and its published/modified date, if shown:
Date source was checked:

Starting facts or accepted base image identifier:
Complete relevant card fields / instructions / Persona / memory snapshot:
Settings before change (unavailable is not zero or default):
One thing being changed:
Field or control being used:
Batch size, duration or other charged option, if applicable:
Displayed cost before submitting:

Exact message or prompt:
[Paste your submitted input here.]

Expected result and specific pass/fail criterion (set before running):
Actual submitted request / attempt identifier, if shown:
Observed reply or media identifier and retained copy:
Actual result:
What worked:
What failed:
What could not be assessed:
Accepted output or timeline version:
Displayed charge or balance change, if observable:

Decision: Keep / Revise / Return to previous version / Stop and investigate
Next change to try:

For state tests:
Value before:
Completed event and its ID:
Value after:
What the narrative says (may differ from displayed tracker state):
Was this a new event or a retelling?
Did an edit, regeneration, or continuation affect the record?
Which version was retained, and what dependent records were reconciled?

For image/media tests:
Source image or clip identifier:
Requested visible change:
Identity / layout / wording preservation criteria:
Unintended changes:
Unassessable details (for example, eyes too small to judge):

Leave unavailable settings unspecified. Never invent a seed or hidden parameter.
A planned expectation is not an observed pass; one result is not a success rate.
Check a pending request's status before retrying a potentially charged operation.
Keep private chat/account data local; redact before sharing a support example.
For prompt-wording comparisons:
Condition: original baseline / revised wording / explicit reminder / continuation repair
Exact original and revised payloads, each destination and lifetime:
How card, Greeting, Persona, pins, summary and settings were restored:
Inputs that could not be matched:
Ordinary baseline message (without reminding the desired behavior):
Separate reminder/stress message, if used:
Attempt count planned before judging results:
Every retained attempt, regeneration and failure:
Intended criterion and unintended drift assessed separately:
For a language trial: requested language, observed language, identifiers retained,
and any part of language quality the reviewer could not assess:

Focused memory, group and tracker checks:
Use Native_Creator_Test_Log.csv for MEM01/MEM02, GR02/GR03 and ST01/ST02/ST03.
Duplicate a case row for each actual attempt; keep exact inputs and full outputs in local records.
Record memory before/after, selected respondent versus fictional turn owner, and native tracker display versus narrated state separately.
The initial rows are proposed checks marked not_run, with observation and verdict fields blank.

For supplied-source and changing-state tests:
Baseline character revision and every active field/preset recorded:
Clean baseline definition and uncontrolled differences:
Source image -> crop -> each attempted edit -> accepted result identifiers:
Displayed charge for every attempted pass; failures retained:
Native generation versus external editing, splicing or audio work:
Current form, stage, accepted transition event and unchanged anchors:
Clothing coverage and features that cannot be judged:
Tracker evidence window and one qualifying accepted event:
Raw classified delta / applied capped and clamped change:
Defined fictional day and gross positive gain already applied:
Physical roster / remote channels / established outfit evidence:
Scenario access, separate character knowledge and optional choice inspected: