Skip to main content

Task Chat

The Task Chat is the conversation that lives next to a single asset in the epic editor — one post, one email, one ad, one banner. It's where 80% of your time inside an epic actually happens. You generate, then refine. Glowtify nudges, you push back, you converge.

If you're working at the campaign level (brief, channels, objective), you're in the Campaign Chat — a different surface in the same page. For the patterns common to every chat in Glowtify (input, suggestions, attachments, MCP), see Working with the AI chat.

Where it lives​

When you select a task in the task list — for example "Facebook post" inside a campaign epic — the right side of the page switches to the Task Chat for that task. The previous Campaign Chat history stays put; you can switch back and forth.

The Task Chat is task-scoped: it knows the task's channel, brief, current content, history, references — and the chat replies are shaped by that context.

Output mode (Text / Image / Text+Image)​

This is the most important control in the Task Chat, and the one users miss most often. Before generating, pick what the chat should produce:

  • Text only — generate or edit only the copy (caption, body, subject…), leave the visual alone
  • Image only — generate or edit only the visual (regenerate, retouch, recompose), leave the copy alone
  • Text + Image — full regeneration

The output-mode toggle sits at the top of the chat input. If your edit only meant to touch the image and the chat rewrote the caption, you likely had Text+Image selected — switch to Image only and re-run.

tip

The output mode is the single biggest source of "the AI changed something I didn't want" confusion. Default behavior is Text + Image on first generation, Text only on follow-up refinements — but you can override on every message.

Target edits​

If you want to change something specific (the headline, the CTA button, the second sentence, the person in the photo), use a target edit chip. The chat surfaces clickable chips above the input — Edit headline, Change CTA, Replace image — that pre-scope your next message to that target. Glowtify will only touch what you targeted.

This is the trick for not re-rolling the whole asset every time. If you don't see a chip for what you want, just describe it precisely in text: "Only change the second bullet, keep everything else identical".

Reference images​

You can attach up to 3 reference images per task to anchor the visual style. Drag images into the chat input, or paste from clipboard. Glowtify uses them as style guides, not as 1:1 sources.

If your task involves a multi-image collage or mosaic that needs more than 3 references, that's a current limit — workaround is to brief the composition in text and let Glowtify generate the layout.

Suggestions and per-action tools​

Below each reply, the Task Chat surfaces suggestion chips — context-aware next actions. They depend on what just happened:

  • Right after generation: "Make it shorter", "Try a different angle", "Add a CTA"
  • After an image regen: "Brighter mood", "More product focus", "Less text overlay"
  • When the asset is close to final: "Schedule it", "Send for review", "Boost this post" (Meta posts only)

Each suggestion is a one-click message — no typing required.

What the chat can do beyond editing​

The Task Chat is more than copy/image refinement. From the conversation you can drive real changes on the task:

  • Schedule the task — set the publish date inline
  • Send for review — if your workspace uses Delegated review
  • Generate variants — "Give me 3 versions" and the chat splits the task
  • Pull from MCP apps — "Use the latest 5 leads from HubSpot to draft a re-engagement email". See External tools (MCP)
  • Attach files — drag a PDF, image, or doc; the chat ingests it as additional brief context

Task Chat vs Campaign Chat​

The split matters because each chat knows about different things:

Campaign ChatTask Chat
ScopeCampaign-level brief and structureOne specific task
What it editsBrief, channels, objective, personasThe asset itself — caption, image, ad copy
Power-up granularity"Add Instagram", "drop email""Edit headline", "make image brighter"
Output modeNot applicableText / Image / Text+Image — most important control

Common workflow: brief in the Campaign Chat → generate tasks → click into a task → iterate in the Task Chat.

What it doesn't do​

  • It doesn't edit the brief. Brief changes happen in the Campaign Chat or the brief panel.
  • It doesn't restart the task from scratch. To redo from zero, use the "Regenerate" power-up in the menu, not the chat.
  • It doesn't see other tasks in the campaign. Each Task Chat is scoped to its own task — for cross-task coordination, go up to the Campaign Chat.

Common issues​

The AI claims it edited the image but the image looks the same​

Known issue — the chat sometimes reports success on image edits that didn't actually apply. Refresh the task or click on the image to force-reload. If the image still hasn't changed, retry your prompt with more specific direction.

My uploaded reference image isn't reflected in the output​

Known issue — reference images aren't always honored. Try rephrasing your prompt to mention the reference explicitly: "Use the attached image as a style guide for color palette and composition".

The chat refuses to edit the image mid-task​

Known issue — for certain task types the image pipeline isn't available. The chat will respond with text only and explain the limitation. If you need image editing for that task type and it's blocked, contact support.

"That didn't go through" on a newsletter generation​

Known issue — newsletter (Beefree) editor occasionally times out on chat-driven edits. Retry once. If it persists, edit directly in the Beefree editor and use the chat for copy iteration only.

Logo removal request errors out​

Known issue — image edits that involve removing or modifying a logo systematically error. Workaround: regenerate the image without the logo in the brief, rather than asking to remove it after the fact.