Can one successful attempt prove repeatable extension quality?
No. The ledger documents that attempt under its recorded inputs and account state. Broader repeatability requires a separately designed evaluation.
Seedance 2.5 Extend / tools / task entry
Seedance 2.5 Extension Continuity Generator turns focused inputs into polished creative results.
Prepared workflow
Prepare a shot-extension ledger for boundary frames, invariant cues, prompt versions, join defects, terminal frames, and editorial decisions.
The source tab needs shot ID, asset version, boundary timecode, extracted frame link, source duration, measured properties, rights status, action phase, camera vector, focus state, lighting note, subject anchors, props, and scene geography. Add a protected-versus-flexible flag to every continuity cue. The ledger should reject an attempt when its source cannot be traced or when the boundary changed without a new record. This avoids comparing outputs from different starting states and calling the difference model variation. Product names may identify the research query, but the source record must describe actual files and observations.
Assign each attempt a version, operator, timestamp, authorized tool and account context, input assets, prompt hash or stable text link, settings actually visible, requested duration, and output file. Keep unverified expectations in a notes field, never in observed-settings columns. Score the join for identity, anatomy, prop continuity, geometry, texture, camera motion, exposure, action phase, and background persistence. Score the terminal frame for editability and next-shot compatibility. When access, controls, or limits differ from the plan, record the discrepancy rather than retrofitting the worksheet to imply success.
For each attempt, choose accept, trim, repair, cover with cutaway, regenerate, switch method, or reject. Require a timecoded reason, reviewer, next owner, and target retest. A comparison view should keep the same rubric across attempts and surface whether a higher visual score introduced a worse ending or provenance gap. Link only the approved output into the edit project and preserve rejected evidence long enough for the production policy. The ledger documents a bounded test; it does not generalize one result into a claim about current product quality, availability, or repeatability.
Capability boundary
Dated evidence
Product and model details can change. These links identify the evidence checked for the claims scoped below.
Circulation only during the recorded collection window; not product, capability, quality, access, or outcome evidence.
Reader notes
No. The ledger documents that attempt under its recorded inputs and account state. Broader repeatability requires a separately designed evaluation.
Public discussion can identify a useful topic, but it does not verify current product features, access, quality, repeatability, or results. Confirm material claims with current first-party sources and the authorized workspace.
Prepare a Wan 2.0 text or first-frame video brief, choose delivery constraints, and hand the structured prompt directly to the current Wan template in SEELE Generation.
Open field notePrepare a shot-level character checklist for reference versions, protected identity cues, scene state, drift defects, and approval status.
Open field notePlan a controlled cloud-generation intake with rights, account scope, job inputs, output custody, review evidence, and explicit stop conditions.
Open field notePrepare a campaign review board for character rights, asset provenance, claims, disclosure, channel rules, defects, approvals, and incidents.
Open field noteContinue the workflow
Verify only the current visible workspace context relevant to this reader task.