Automations

Automation Engine

The Automations page supports both individual automations and connected workflows for the active Synfluencer. A workflow runs discovery, content creation, draft building, media generation, scheduling, and publishing in order under one schedule. Exact snapshots and item-level lineage keep titles, platform versions, media, approval state, and run history traceable end to end.

How to access

On desktop, choose Automations in the sidebar, directly below Review. On mobile, open More and choose Automations. Requires an active Synfluencer and Editor or Owner access.

Automation types

• Trend Discovery — Finds and deduplicates current trends, then records each title, source, and Persona relevance score in the run.
• Trend → Content — Selects the highest-relevance trends from the connected discovery run and creates content through the same Persona, Brand, Voice, creative-system, intent, and platform-capability rules used by Create and Discover. Choose an intent or leave it on Auto. Choose quality-gated automatic approval (80/100 by default), always approve and continue, or review every content piece. The policy applies to newly generated content; changing it never retroactively approves an existing Review item. Blog, Newsletter, Hashtags, Video script, and Source link are optional and start off. Source link appends the primary source URL to supported posts while X remains link-free.
• Draft Builder — Turns both Review and Approved content from the connected run into exact platform drafts without generating a second content piece. It inherits the saved intent and optional additions, prefers the selected A/B version when enabled, and updates an existing editable platform draft instead of duplicating it. A required image or video is handed to Asset Generation; it is not a Draft Builder approval gate. Approval is still required before scheduling or publishing.
• Asset Generation — Runs after Draft Builder because it needs each exact platform version and its saved media rule. It creates images, audio, or one selected video outcome and attaches each result back to those drafts. A selected identity photo stays pinned to that content piece for stills and video; X and similar feeds use landscape stills, Instagram and other static feeds use square stills, and 9:16 remains a video frame.
• Scheduling — Assigns eligible Approved drafts through the shared audience-local recommendation engine. It can inherit optional profile publishing hours or use its own 12-hour AM/PM window; existing posts count toward spacing and local daily caps.
• Publishing — Submits due Approved, Scheduled drafts to connected platforms after required-media checks pass.

Build a connected workflow

1. Choose Create workflow. The button is available even when no individual automations exist.
2. Name the workflow, add two to six automations, and configure each block inside the composer. You can optionally reuse an existing automation, but it is never required.
3. Drag the blocks into pipeline order and set the workflow's one daily, weekly, interval, or manual schedule. Invalid or duplicate automation orders are rejected before save.
4. Review and save once. The workflow and any new automations are created together; canceling leaves no partial automations, and saving does not run providers or spend credits.
5. Use Preview workflow to compare exact currently stored inputs with projected handoff capacity. Projection is a configuration limit, not a promise that every automation will produce that many items.
6. The Run order cards report each automation's place and current status; they are not selected before a run. Choose Edit workflow to change the pipeline.
7. Use Run workflow for one immediate sequential run. Enable schedule controls future saved times and never starts or restarts the current run. When a run is waiting, Resume from [automation] names the exact stage it will retry, keeps completed stages, and does not resubmit provider work that is still processing. A second full run remains unavailable until the open run ends.
8. Follow progress without refreshing: the active automation alone shows a spinner and Working, earlier automations stay Completed, and later automations say Waiting. Action needed appears only when a real approval or setting blocks the run. Last run and Runs include active attempts instead of waiting for the whole workflow to finish. After full success, the workflow header keeps Completed while compact block pills clear; expanded run history keeps every status. After a failure, Continue from failure reuses completed automations and starts a separate continuation at the failed automation.
9. Open Recent runs and choose the exact dated run you want to inspect. The newest run opens first and only one run is shown at a time; load older runs in bounded groups when needed. Every completed stage offers Rerun from this automation, even when a later stage in that source run is still working or waiting. Choosing the first stage reruns the full pipeline from that run; choosing a later stage carries its earlier output forward exactly and reruns that stage plus everything after it as an isolated comparison. Existing copy, drafts, media, schedules, and run history remain current during generation. Publishing is previewed but never submitted at this point. Review both attempts, then explicitly Keep updated version or Disregard changes. Only one comparison can be open at a time; Disregard leaves an open source run unchanged, while Keep closes only its unfinished stages before promoting the candidate.

Individual automations remain useful for one-off recurring jobs. Run connected chain runs an existing dependency chain immediately without turning it into a saved workflow. Automation workflows are separate from Create → Video recipe workflows: they connect product automations; video recipes create one media outcome.

Approval and paused-stage recovery

Quality-gated automatic approval keeps content at or above the saved score moving and sends lower-scoring content to Review with its score and threshold. Always approve and continue is an explicit owner choice that removes the score stop for newly generated content; Review every content piece always waits. When Scheduling or Publishing is waiting, expanded history groups repeated platform rows under one content title. Choose Review and approve to inspect one complete Draft workbench, or Approve and continue to confirm the exact grouped batch. Approval automatically wakes the same workflow run from its waiting stage. Approval alone never publishes content.

For settings, consent, spending, or required-media checkpoints, save the correction and choose Resume from [automation]. The workflow retries only the named stage and records the previous stage attempt for lineage. If the stage is still generating provider work, Resume reports that it is still working and leaves it alone. A failed stage uses Continue from failure, which is a separate recovery path and run.

Media behavior

Draft Builder uses one clear media rule per platform: Text only, Text first, Use image if available, Require image, Use video if available, or Require video. Asset Generation reads those saved rules, generates only media a matching draft can use, and attaches it to the exact draft. There is no random “image chance” and no conflicting Prefer image / Prefer text state. Selecting Image or Video in Asset Generation does not override Draft Builder: the editor and Review step warn when the connected draft policy would skip that asset, and a real run records the affected title, platform, saved policy, and recovery action instead of reporting an unexplained zero. Unless the saved prompt explicitly asks for fantasy or stylization, generated stills default to grounded photorealism, plausible anatomy, natural light and materials, and the selected person's visible identity and wardrobe. A fantasy or stylized request remains authoritative.

Customer Asset Generation supports three direct tools: Talking video, Generate motion, and Adapt video. The six former guided recipes remain available only to admins for retained testing and history. Eligible talking outcomes use one model: Avatar IV, LC Avatar, MX Avatar, or Aurora Avatar. Narration uses the profile's explicitly saved ElevenLabs, Gemini, or xAI catalog voice or private authorized Fish Audio clone. UGC Ad keeps the real product Asset separate from the presenter image; choose an approved product Asset or explicitly generate an original, source-grounded product concept. Every selected presenter reference requires likeness permission; source video and cloned voice keep separate rights confirmations. Turn on Automatically approve narration to save the generated track to Assets and continue directly to video. Leave it off to listen and approve it in Speaking voice first. If no primary voice is saved, the run stops before provider spend and links to the profile chooser. One-off and manual Video work keeps its normal narration review step.

Open Asset queue from an expanded workflow run to see each content piece once, its platform destinations, live image/audio/video previews, and whether it is waiting, working, completed, or genuinely needs review. The queue scrolls as one surface, so its title never covers later assets. Background work needs no click and continues after you leave the page. A manual narration or guided-video checkpoint links to that exact review; configuration blocks link back to Asset Generation. Finished media opens with the complete post in Review, where you can approve, replace, or regenerate it. Completed talking videos pass a full decode, audio-presence, and black-frame check before they can attach; an unusable provider result fails visibly and releases its reserved charge instead of becoming a black preview with audio.

Managing automations

• Create — Set up a new automation by selecting the type and configuring the schedule. A successful save closes the editor; any schedule warning appears as a toast on the updated list.
• Enable / disable — Toggle automations on or off without deleting them.
• Run manually — Trigger an automation immediately outside its schedule.
• Reset — Clear the automation state and start fresh.
• Preview run — See the exact current items that would process, skip, or need attention without calling an AI provider, reserving credits, creating drafts, or publishing.
• Recover a skipped run — A no-input result is labeled Skipped and links directly to Run Trend Discovery first or Run connected chain. If generated work was later deleted, Rebuild deleted content or Rebuild missing drafts restores only the missing output.
• Control a workflow — Enable schedule and Pause schedule affect future saved times. Run workflow starts the full ordered pipeline once. If a run is already open, its named resume, review, or comparison action replaces guesswork and a second full run remains unavailable.
• Continue a failed workflow — Status pills identify the failed automation. Continue from failure asks for confirmation, keeps the failed attempt in history, reuses earlier completed output, and reruns only the failed automation and those after it.
• Compare a rerun — Select a dated run, then start from any completed automation inside it, including one whose later stage is still open. The comparison uses that run's exact lineage, regenerates existing assets instead of treating them as a reason to skip, and waits for Keep or Disregard before replacing anything or releasing Publishing.
• View history — Expand recent runs to follow an active run as it refreshes. The newest run is visible by default; a dated selector switches the one displayed run and Load older runs retrieves more history without lengthening the page. Draft Builder groups platform drafts under their one content title; Scheduling and Publishing group approval waits the same way and expose individual or batch approval. Asset queue shows media previews plus exact draft, narration, settings, or video-workflow recovery links. Counts name the automation output—trends, content pieces, platform drafts, or asset jobs—so platform versions are not mistaken for duplicate content.

Trend reuse and minimum relevance

Trend → Content consumes individual eligible trends, not an entire Discovery run at once. If a discovery found ten eligible trends and one content run used three, later runs can use the remaining seven while they are still inside the configured lookback period. Minimum persona relevance is the pre-generation audience-match score for a trend; it is not the generated draft's QC score. A value of 60 means trends below 60 are skipped before content generation and remain explained in Preview and run history.

Premium X and Long Posts

Trend → Content reads the same connected X Premium setting used by Create and Discover. When enabled, X generation produces one Long Post, keeps the strong first 280 characters before “Show more,” and Draft Builder preserves that Long Post instead of converting it into a thread. An explicitly created thread remains a thread. When Premium is off, Draft Builder can safely turn copy over 280 characters into a thread.

Credits and spending ceilings

Every AI action created by a scheduled or manual automation appears in AI usage with an Automation label and the automation name. Its billed amount contributes to the same profile monthly total, Home Spend this month, billing summary, and projected spend as interactive generation. Charges are recorded per AI action, so completed discovery or provider-reported generation usage still updates spend when a later automation or QC step fails. A fresh customer estimate is calculated before paid media starts. The saved safety maximum is a hard stop by default; if an estimate is higher, the item pauses and shows the exact amount required. Choose Update spending ceiling, save the change, then choose Resume from Asset Generation to retry that same paused stage. If you explicitly turn on Continue above this maximum when existing credits cover it, the current bounded estimate may continue above that amount. Available credits are always checked, so the override never creates credit or permits an unaffordable generation.

Audience-local scheduling recommendations

Scheduling uses the same recommendation system as Bulk Schedule and Calendar Autofill. It scores exact post history in the Synfluencer audience timezone, can broaden sparse weekday evidence to the same local hour across days, and only then uses platform guidance interpreted directly as an audience-local clock time. Turn on Use profile publishing hours to inherit the optional Brand profile window. If that profile window is not saved, engagement history and platform recommendations stay available without adding an artificial restriction. Turn it off to set an automation-only earliest and latest time with 12-hour AM/PM controls. After scheduling, Calendar shows the audience-local time while the database and provider handoff keep the exact UTC instant.

Timezone

Schedules fire in the active synfluencer's audience timezone. Choose each run time with an explicit 12-hour AM/PM control: for example, 1:00 PM in EST is stored and fired as 1:00 PM EST. A “Daily at 9 AM” automation on an EST-audience synfluencer fires at 9 AM EST, while the same rule on a PST-audience synfluencer fires at 9 AM PST — each synfluencer owns its own clock. Change it in Settings.