Start for free
Browse docs▾

Chat and generation

How to ask the AI for drafts, attach references, control video length, add captions or dubbing, then edit and approve before publishing.

The chat is your main workspace. Describe the outcome in plain language. The assistant drafts copy and media, then you refine until the post is ready.

How to ask for good results

Strong requests usually include:

  1. Platform (LinkedIn, Instagram, TikTok, X, Facebook, Pinterest, blog)
  2. Format (single post, carousel, Reel, Story, Short, thread, pin, article)
  3. Topic and goal (announce, teach, inspire, drive traffic)
  4. Tone and language
  5. Constraints (length, no hashtags, include a CTA, avoid jargon, target duration)

Example prompts:

  • “Draft a LinkedIn post announcing our webinar on March 12. Hook in the first two lines. Professional but warm. End with a clear CTA.”
  • “Create a 7-slide Instagram carousel on morning routines for founders. 4:5, bold on-slide titles, friendly caption.”
  • “TikTok video, about 20 seconds, hook in 2 seconds, on-screen captions, about inbox zero habits.”
  • “Write an X thread (8 tweets) summarizing our report. First tweet must stand alone as a hook.”
  • “Add burned-in subtitles to this video.” / “Dub this Reel into Spanish.”

You do not need to name AI models. Describe the result; the assistant chooses how to generate it. See Choosing models.

Attachments and references

You can attach files in chat (paperclip button):

  • Up to 6 files per message
  • Up to 100 MB per file
  • Images, video, and audio are supported for generation workflows

How the assistant uses them:

  • Reference image: steers the look of a new still or the first frame of a video
  • Reference video: extracts storyboard frames (for example first, middle, last) so image-to-video and follow-up creatives stay consistent with your clip
  • Voiceover audio: can drive clip length and be muxed onto generated video; long voiceovers can be split across several clips that are then merged

Say what you want done with the attachment (“use this as the first frame”, “match this style”, “add captions to this file”).

Video length

Platform tips often mention Reels or TikToks around 15 to 60 seconds. That is still good creative advice.

In auto-social.io:

  • Ask for a duration in plain language (“about 8 seconds”, “around 20 seconds”, “about a minute”).
  • Each single generation is snapped to what the chosen video model supports (often a few seconds up to roughly 15 seconds per clip; some models allow a bit longer).
  • For longer videos, the assistant can generate several clips and merge them into one file. This is especially useful with a longer voiceover: the audio is planned into segments, each segment is generated, then the parts are combined.

You do not manage segment indexes yourself. Describe the total length or attach the voiceover and say you want one finished video.

Captions, subtitles, and dubbing

Ask in chat when you need:

  • Subtitles / captions from spoken audio or an uploaded video: speech-to-text exports .srt and .vtt. You can also ask to burn captions into the MP4.
  • Dubbing / re-voice into another language (audio translation). This is separate from captions and uses more credits (on the order of thousands of credits per minute of source). Only ask when you explicitly want a translated voice track.
  • Voiceover or multi-speaker dialogue on a cinematic clip, or a talking-head / UGC face-cam style when you want a presenter’s face animated from a still.

Generated subtitle files and media also appear in Library.

Drafts

Generation creates drafts, not instant live posts (unless Autopilot is enabled).

A draft may include:

  • Caption or body text
  • Hashtags or alt text when useful
  • Images, slides, or video assets
  • Notes or structure for multi-part formats (carousel, thread, document)

Treat the first reply as a strong starting point. Most great posts come from one or two short refinement rounds.

Edit and refine

You can refine in chat or edit the draft fields directly.

Useful follow-ups:

  • “Make it 30% shorter.”
  • “More concrete examples, less fluff.”
  • “Rewrite for French, keep the structure.”
  • “Swap the image for something brighter and more lifestyle.”
  • “Keep the caption, regenerate only slide 1.”
  • “Make the video about 25 seconds with captions.”
  • “Add a soft CTA without sounding salesy.”

Be explicit about what to keep and what to change. That reduces credit spend and avoids losing good parts of a draft.

Approve, schedule, or publish

In Assistant mode:

  1. Review copy and media.
  2. Edit until it matches your brand.
  3. Approve.
  4. Publish now or schedule for later.

Nothing goes live without your approval in Assistant mode. Autopilot is separate and optional. See Assistant vs Autopilot.

Formats and networks

Each network has preferred ratios and habits. Asking for the right format up front improves results:

Languages

Ask in the language you want, or specify it (“caption in Spanish”, “thread in German”). Copy can be prepared in 90+ languages. Spoken voiceovers cover 32 languages.

Common mistakes

  • Vague prompts (“make a post”) without platform or goal
  • Requesting video before the message is clear
  • Mixing carousel ratios across slides
  • Using landscape creatives for vertical-first networks (TikTok, Reels, Shorts)
  • Expecting one model call alone to always be a full 60-second clip (ask for the total length; longer cuts are built by merging clips)
  • Publishing without a final read for brand names, dates, and links

Related