How AI model selection works
The auto-social.io assistant picks the best model for your request. You describe the outcome; you never manage model ids.
The auto-social.io AI assistant picks the best model for you based on your request and need. You do not need to name models, compare providers, or manage model ids.
Describe the outcome, not the engine
Tell the assistant what you want:
- Platform and format (LinkedIn post, Instagram carousel, TikTok video, blog article)
- Quality feel (quick draft vs polished final)
- Media needs (text only, with images, with video, with or without voiceover)
- Clip length when video matters (“about 8 seconds”, “around 30 seconds”)
- Tone, language, length, and constraints
Examples:
- “Quick draft caption for a LinkedIn update, text only.”
- “Polished Instagram Reel with on-screen captions and a short voiceover, about 20 seconds.”
- “High-quality 7-slide carousel, crisp product photos, friendly caption.”
- “Talking-head UGC clip from this portrait photo.”
- “Add subtitles to this video” / “Dub this video into French.”
The assistant chooses the generation approach that fits that need (text vs image vs video, format, quality preference, and length).
Always current models
auto-social.io uses the latest AI models available, kept current in the product. You benefit from up-to-date capabilities without tracking releases or swapping settings yourself.
You will not find a shopping list of internal model catalogs in this documentation. That is intentional: model choice is handled for you.
Optional levers (plain language only)
You can influence results without naming engines:
| What you say | What it helps |
|---|---|
| “Quick draft” / “rough first pass” | Faster, lighter starting point |
| “Polished” / “campaign-ready” | Higher-quality finish |
| “Text only” | Avoids unused media spend |
| “With one image” / “7 slides” | Right media pack size |
| “With voiceover” / “no voiceover” | Audio when you need it |
| “About N seconds” / “around a minute” | Target video length (see below) |
| “With captions / burned-in subtitles” | Speech-to-text files or on-video captions |
| “Dub into Spanish” (or another language) | Translated voice track (higher credit cost) |
| “More natural” / “more bold visuals” | Creative direction |
| “Use this image / video as reference” | Attachment-driven look or first frame |
Higher plans may expose quality preferences in the product UI. Even then, you are choosing an outcome preference, not picking model names by hand.
Video length without picking a model
You never choose which engine supports which duration. You only describe the result:
- Ask for a target length in seconds or a short phrase (“short hook”, “about half a minute”).
- The product snaps each single clip to durations that engine supports (often up to roughly 15 seconds per generation; some options go a bit longer).
- For videos longer than one clip, the assistant can generate multiple segments and merge them into one file. Long voiceovers are a common trigger for this multi-clip plan.
Details and attachment tips: Chat and generation.
What you never need to do
- Paste or memorize model ids
- Compare internal catalog strings
- Manage provider accounts for generation inside auto-social.io
- Guess which engine is “best” for a Reel vs a LinkedIn post
- Manually stitch clip indexes when you asked for one longer video
If you are unsure, just describe the post. The assistant will route generation correctly.
Why this design
Social content mixes text, images, and video. The right engine depends on modality, format, and quality bar. Letting the assistant choose:
- Saves time
- Avoids mismatched settings
- Keeps your workflow focused on message and brand
For how to write strong requests, see Chat and generation. For cost of generation, see Credits and plans.