AI rewriting for Telegram posts
Let a language model rewrite, summarize, translate, tag or filter every post before it reaches your Telegram channel. Ready-made and own prompts, own key.
@AximoBot can pass every incoming post through a language model before delivery. The model works on the text of the post, so the same stream of sources can arrive as short summaries, reworded news, translated posts or hashtagged entries — different for each chat or even for each source. AI processing is on the Expert plan and above; the platform pays for the models and there is no token limit.
What the model can do
The prompt library has ready-made prompts you switch on with a tap:
- Translate the post into your language, Retell in different words, Bring it to a business tone;
- Short bullet-point digest and Pick hashtags, Detect the topic and tag it;
- Cut promo lines and links to other channels and Tidy up the text formatting;
- Strip emotion and opinion;
- Tag promotional posts — marks ads so a filter can drop them;
- Break a trading signal into fields and Mark how urgent a trading signal is.
The prompts that are switched on are combined into one request per post and run top to bottom; you can drag them to reorder.
Your own prompt
A prompt of your own is free text, in any language: "rewrite as a two-sentence news item, keep numbers and names, no emoji" or "translate to German, keep brand names in English". It is saved already switched on and always rewrites the text. Before trusting it, use Prompt test: paste a post or load the last one, press Apply prompts and see the result with the same settings a real post gets — your prompts, translation, filters and model.
Where it is set
For the whole chat or for one source: a source with prompts of its own either replaces the chat's shared prompts or adds to them. A chatty source can be turned into a short summary post with a summary prompt plus the First media only switch in Message content. See AI processing for every screen.
Filtering by meaning
The same model can decide what not to publish. A prompt marks the post (for example the ready-made ad tag), and a filter rule at the After AI stage drops the marked ones — AI filters. This catches ads and off-topic posts that a keyword list cannot.
Where AI sits in the pipeline
AI runs after replacements and translation, so prompts see the already cleaned and translated text; "After AI" filter rules run right after it. The model does not see the media — only the text — with one exception: for YouTube sources the Video subtitles to AI switch adds the video's transcript to the model's context, so a summary reflects what is said in the video. The model's answer is tidied for Telegram (headings become bold, lists get bullets, tables become plain lines); Disable post-processing keeps it exactly as written, for example for code. The full order is in A post's path.
Your own key
By default the bot uses its shared OpenRouter account and a model it picks. If you want to choose the models and pay for them yourself, paste your own OpenRouter key in LLM settings: the first model you pick is primary, the next ones are fallbacks. If none of your models answers, the shared pool processes the post and you are notified; a key OpenRouter rejects is cleared and the owner is told.
Questions
Does AI work on the free plan?
No. AI processing starts on Expert. Automatic translation works on any paid plan (Pro translates with Google, Expert and above with an AI model).
Will the model change the language of my posts?
Only if the prompt says so. The ready-made prompts ask the model to keep the language of the post; if you need a specific language, name it in the prompt. The answer's alphabet is not checked.
Can it handle photos or video?
No: it works on text. The media travel with the post unchanged, except that a watermark or logo removal (Ultra) can be applied to photos separately.
How do I keep a post unchanged when the prompt does not apply?
Several ready-made prompts only act on matching posts (an ad tag, a trading signal) and otherwise leave the post as it is, while the other active prompts still run.