ZimmWriter Changelog

Released 2025-08-09

• Ollama support is now integrated in ZimmWriter for 100% FREE AI unlimited usage!
  • Four slots are added to ZimmWriter’s Text API area for these models.
  • Now you can run locally hosted models.
  • Note that your performance will vary and many local models suck, which means ZimmWriter won’t even finish writing an article. So if you’re using a locally hosted model, PLEASE PREFACE ANY FACEBOOK POST WITH THIS FACT.
  • In addition, you’ll want to adjust the context length to at least 8k in the Ollama settings.
  • Ollama “turbo” is not supported, so just use the free version which has no limits.
• Fixed a bug that would prevent a shortcode from ending a section (e.g., via a raw prompt) when using a linkpack.
• Fixed a bug where some apostrophes + s (so ‘s) were turned into ‘S in titles.
• Fixed a bug that would cause ZimmWriter to crash when an AI image provider generated an image but it was zero kb in size (so basically a glitch by them).
• Added the ability to access {varX} inside of raw prompts and custom prompts.
• Boosted the number of {varX} available to fifteen, so {var15} works!
• Added Seedream 3 to Runware as an image model.
• Added Flux Krea to Runware as an image model.
• Upgraded Claude Opus 4.0 to 4.1 (but it will still show 4).
• Added GPT-5 variants but it’s not very good for actual writing. It’s decent for ancillary tasks though.
  • All varients added to OpenAI’s ZimmWriter integration.
  • Mini and Nano addded to OpenRouter’s integration. I cannot add GPT-5 because at the moment OpenRouter requires you to use your OpenAI API key which defeats the purpose. So if you want to use it, you can use it as a user choice model for now.
  • Added an O button to some of the writing model dropdowns to select the reasoning level. By default, reasoning is set to minimal. Though for some odd reason, OpenRouter doesn’t have minimal and only has low.

Released 2025-07-12

• Added Grok 4.
• Bug fixes to the social media scheduler.

Released 2025-07-06

• Fixed a bug that would cause amazon images from AFFI to be pulled when using the social media scheduler after editing the post.

Released 2025-06-19

• Patches a JSON error that would occur for some people when ZimmWriter was communicating with the AI.

Released 2025-06-02

• Patches a JSON error that would occur for some people (rare situations) when uploading to wordpress.
• Added the ability to backdate WordPress posts 25 years in the past.
• I removed Gemini because Google doesn’t have their act together. I was testing out Flash 2.5 and their Pro 2.5 and I was getting multiple failed replies, but still charged. Google is a mess when it comes to AI via API. Stick to other providers.

Released 2025-05-26

• Overhauled listicle detection and processing so that it’s now much more accurate. It fixes a lot of previous issues, such as situations where you’d have a non-listicle number at the start of a title (e.g., 2100mha Battery Review). Also, just to clarify, a listicle number overrides the “# of H2” setting and “H2 Auto Limit” setting, however the listicle cap remains at 50. It has always worked this way, but I think adding this clarification every now and then is helpful.
• Added Hi-Dream AI image models to Runware (RW).
• Added Google’s Image Gen 4 to Fal (F) and Replicate (R).

Released 2025-05-22

• Fixed a bug that would cause assignment of categories with “&” symbols to glitch out when uploading to WordPress.
• Added Claude 4 Sonnet and Opus. I replaced Opus 3 but left Sonnet 3.5 and 3.7.

Released 2025-05-16

• Removed all instances of Ideogram 2 (except for Replicate’s 2at due to cost) because Ideogram is going to discontinue the model. So PLEASE update all your profiles in the Bulk Writer to one of the Ideogram 3 models.
• Added Ideogram 3 quality, balanced, and turbo models.
• Added a random string to end of image file names when uploaded to WordPress to prevent duplicate filename issue which could lead to non-uploads and re-used images due to server cache.
• Patched a bug with non-English slugs. I think they should work better now. Let me know.
• Added some patches to the AI outline generator to prevent some models (such as Claude Sonnet 3.7) from leaking the title into the outline.
• Boosted the length of a {varX} from 50 characters to 100.
• General bug fixes.

Released 2025-04-25

• Boosted list of possible do-follow domains in the link toolbox to 1000.
• The product roundup feature now retains the list of failed plural product names and URLs.
• A bunch of fixes the “up to 5 non-link pack links” feature. I don’t use it much so it wasn’t until I started to test it out heavily did some bugs appear.
• Improved bulk slug generator.
• Big improvement to reasoning models for Sonar.
• Fixed a bug that would cause missing featured images when zimmwriter renamed a subheading image for SEO purposes and the renamed image shared the same name as the featured image file.
• Added the new gpt-image-1 model by OpenAI. But…
  • 1. You need a photo identification (e.g., your license or passport) to use it (see the link below).
  • 2. You’ll need an OpenAI API key (it only works via their direct platform).
  • 3. You can easily hit lots of “safety” hiccups preventing the image from being generated depending on your niche. I have no way to give you advance notice of that, so test out your nice first.
  • 4. It’s expensive depending on the quality level you choose. So be prepared, and as always, set monthly spend limits on all your platforms so you don’t get hit with a weird multi-thousand dollar bill.
  • 5. The aspect ratios are fixed (16:9, 1:1, 9:16).
  • 6. Only “high” is good for infographics, while “medium” is okay for light amounts of text.
  • 7. The way text is generated is completely different from how other models work.
  • 8. Your mileage will vary.
  • https://help.openai.com/en/articles/10910291-api-organization-verification

Released 2025-04-21

• Improved handling of nested unordered lists in html and wordpress upload.
• Added a new feature to link packs that allow you to disable insertion of links except via a {lp_ variable at the end of a custom outline subheading. Previously, you could use a {lp_ variable in a custom outline, except that other subheadings would still get links based on the loaded link pack.
• Patched a bug with alt images not applying the prompt.

Released 2025-04-17

• Come to find out that Gemini’s paid model isn’t granting the advertised 1,500 FREE requests-per-day of grounding (i.e., deep research). So I have disabled that model from the research dropdown. I have left the free model for now. I hope Google gets their act together soon. But… I had to rush this patch out in response, and I was working on some features to add to ZimmWriter. Thus, they were not fully vetted by the beta testers. So if you get a crash, then PM me on Facebook so I can debug.

Released 2025-04-14

• One of those days…

Released 2025-04-14

• Found a bug introduced in 10.806 which might trigger unexpected things with the combination of an “alt” image and zero H2.
• Removed GPT 4.5 because who are we kidding.

Released 2025-04-14

• Added GPT 4.1 Mini, GPT 4.1 Nano, GPT 4.1.
• Remove GPT 4 Turbo, GPT 4.

Released 2025-04-12

• Fixed a JSON bug that was introduced in the prior version.

Released 2025-04-11

• The cost in the text file now takes into account image pricing.
• ZimmWriter now adds anchor (jump) links to each subheading for newly created blog posts. What this means is that the social media scheduler will now detect those newly created anchor (jump) links when using subheading images. So when a user on social media clicks on your post, they will be taken to the relevant subheading section in your post.
• The bulk writer now has an “alt” image menu button. You set the model, options, and prompt, and ZimmWriter will generate a second hidden image which only becomes visible when you pin the image or use the social media scheduler. Using the plugin below will also allow a new featured image shared when you share the post.
• New plugin which adds new opengraph headers to your post when you uploaded the post using ZimmWriter’s new alt image feature. The plugin is here (remove the space):
  http://rankingtactics.com/ zimmwriter_og_image_manager.zip
• Fixed a bug that caused ZimmWriter not to translate the H2 in a job with only 1 H2.
• The social media scheduler now includes the ability to select which images you want in the event you’re using alternative images. You can select “main” for non-alternative images, “alternative” for just alternative images, and “both” for both!
• Added an option to the raw prompts to enable “lookaround” mode which feeds in the [PREVIOUS SUBTOPIC] and [NEXT SUBTOPIC]. If you enable lookaround mode, then make sure your raw prompt references [CURRENT SUBTOPIC] so the AI knows what you want it to write about. But the AI will also be fed those additional placeholders for more contextual relevance.
• Replaced Grok 2 with Grok 3 & 3 Mini for OpenRouter.
• Added direct API access for Grok. Note: Grok 3 is a reasoning model so it does a lot of verbose thinking and ends up being kind of expensive. But you can offset this with the free $150/mo in credit if you use data sharing. You can also define the “reasoning_effort” in the O menu.
• Added direct API access for Gemini. At the time of writing this, the FREE model is free with generous limits and gives about 500 free web hits a day via Deep Research. Also at the time of writing this, the non-free model gives you 1,500 free web hits a day via Deep Research (but if you exceed that in a day, then it’s $35 per additional 1,000). NOTE: If you select the non-free model, ZimmWriter has NO way of knowing if you’ve exceeded the free 1,500 or not, so you expect to see the non-free cost relfected in the text file. Yeah, confusing as heck, but you can figure it out. I believe in you.
• Fal’s Flux Realism was removed because it’s discontinued. I have replaced it with some Juggernaut models.
• You now can run custom loras on Runware models.
• For webhooks, the WordPress subheading image URLs are now sent along with any featured alt image URLs and subheading alt image URLs.

Released 2025-04-01

• Bugfixes.

Released 2025-03-27

• As expected, the “BIG” change behind the scenes led to some issues since everyone uses different Windows computers. So I found a solution and now ZimmWriter’s option menu has a “Enable JSON Compatibility Mode” checkbox. By default it is checked. If you’re on the latest version of Windows 10 or 11 and use translations, you could try un-checking this box and seeing if ZimmWriter runs smoothly. Otherwise leave this box checked. What does it do? It modifies how unicode data is sent in JSON to the AI models.

Released 2025-03-26

• The Bulk Writer input box character limit has been bumped to 140,000 characters.
• The featured image prompt will now be fed the original blog post title, with any listicle number intact.
• The featured image alt text is now translated. The image names are also translated.
• Runware models now get many less errors.
• Sometimes subheadings generated by the AI would get a ** added to the end. Now that’s patched.
• The custom background dropdown box and raw prompt dropdown boxes are now wider so you can better see longer names.
• Claude 3.7 is a pain so I added more prompt leak detection, especially to translations, but you still might need to manually fix stuff if you use this model. It leaks more than a baby wearing cheap diapers.
• The ZW WP plugin is no longer needed because non-English characters are sent now to WP as non-English characters, and not escaped. But this brings me to the next point…
• The BIG thing is a behind the scenes change which no longer escapes non-English and special characters. It could lead to JSON errors but I believe I have patched all potential errors. However, with everything, you never quite know until you release something to the masses. So let’s cross our fingers and hope the world doesn’t burn down.

Released 2025-03-08

• Added Claude Sonnet 3.7. Since Sonnet 3.5 is depreciated, I will soon remove it from ZimmWriter. So please update your saved profiles. NOTE: It seems that Sonnet 3.7 LOVES to leak headings. I’ve added a bunch of patches behind the scenes. But if you get duplicated headings, send me a PM with details.
• MAJOR news! The Link Pack Toolbox pipe trick “|” now works in the Link Pack menu Non-Link Pack Links box.
• Added Ideogram 2a and 2a Turbo.
• Added Juggernaut image models to Runware.
• Fixed a bug that would crash ZW when you only put in one link in the up to 5 raw links box.
• WEBHOOK UPDATES:
  • “url” is now passed, but realize if the post is not published then ZimmWriter will try to predict the URL.
  • “date_gmt” is the wordpress publish date in GMT
  • subheading images are now provided, however since they are base64 encoded if you do too many, then the make interface breaks. So I’ve capped the limit on the number that are sent to 5. I have found that this has less of a chance of breaking stuff.
  • “wp_post_id” is passed.

Released 2025-02-16

• Added Runware as an AI image provider. The company hosts open source models and the three models I added are Flux Dev, Schnell, and SD 3. The only reason I added them is their prices are about 80% cheaper! HOWEVER, I have noticed that their API can be flakey, sometimes slow, sometimes returning an error, etc. It seems that because they are new, they are going through growing pains.
• Adjusted {save_text} so that it doesn’t save HTML tags.
• Fixed a bug with scraping surgeon that would crash if the image contained a relative path.
• Fixed a bug in the Penny Arcade where AI images were broken for the featured image. Why does no one ever tell me this stuff?! LOL. Now it’s fixed.
• Image alt text is now translated.
• Wehooks! Learn more here: https://www.rankingtactics.com/zimmwriter-webhooks/

Released 2025-02-16

• Fixed a bug with the Text Discombobulator where {output1} wasn’t being detected in the prompt.
• Added Ideogram to both Fal and Replicate.
  • NOTE: The “Styles” and “Magic Prompt” settings do not work in Fal. I have reported the issue to Ideogram (since they maintain the integration as far as I’m aware). No idea about a time for a fix. However, everything works fine in Replicate (R) and Ideogram (I).

Released 2025-02-14

• Several bug fixes to profile loading using {profile=} and WordPress date scheduling.
• Removed the older Perplexity models from OpenRouter (since they will be discontinued Feb 22 by Perplexity). Also added Sonar Reasoning models to OpenRouter and Perplexity, which are denoted R.
• Other random bug fixes.
• Added a timezone option for each of your WordPress sites in the main WordPress add/delete site menu. As a default, it’s set at UTC+0. It is recommended to set this to your WordPress timezone (in the WordPress settings menu on your site).
• Also here’s something I discovered when writing some news articles. Let’s say you tell ZW to publish 10 articles today and it’s 10am server time. If you set the # per day to 10, then 24hrs/10 = 144 minutes apart (2.4 hours), so ZW won’t be able to fit them all into the current day, so some will schedule for tomorrow. But if you set the # per day to 50, then 24hrs/50 = 28 minutes apart, so ZW will be able to fit them all into the current day. Therefore, when doing a job that you JUST want published TODAY, then make sure your # per day is set to 3-5x higher than the actual number you intend to publish (if you’re starting the job after 12:01am server time). Make sense? LOL.
• Added blockquotes, which when enabled, have a 50% chance of appearing on the first two H2 sections, otherwise 30% on the later H2 sections. The additional conditions are each H2 section must have 4 or more paragraphs + lack a list + lack a table. Also, on ZERO H2 or 1 H2, the rate is 70% (assuming no lists or tables). You can also force a blockquote using {blockquote} using a custom outline.

Released 2025-02-07

• Adjusted the prompt for scraping URLs as it was leading to some failed scrapes when creating a link pack.
• The “temp disable scrapeowl” checkbox in the Link Toolbox is now checked by default.
• Added all kinds of new fun options for the H2 image quantity dropdown
• Added custom prompt functionality to the meta description in the custom prompt menu.

Released 2025-02-05

• The new year revealed a bug in the Social Media Scheduler when using the AI to recommend a date range checkbox. The bug is now fixed and your schedule should be even more awesome!

Released 2025-02-04

• More tweaks to slug generation.
• Fixes to some more JSON errors on WP upload that could occur.

Released 2025-02-01

• The DEEPL paid key can’t work on the free models and the reverse is also true. So to avoid confusion, I’ve added two DEEPL key input boxes. One for a the free account and one for the paid account. You can use both if you want, but you’d need two separate DEEPL accounts (one free and one paid).
• Updated Groq to the latest model.

Released 2025-01-31

• Added DEEPL as a translation model to ZimmWriter. You will need a DEEPL API key. The company has a free tier but you still need to link up a credit card. Note: If you use the paid plan, then the cost is reflected in the estimate inside the text file ZimmWriter generates.
• Added an option to display all references used to generate Deep Research knowledge at the end of the article in a References H2 section. The list displays references, even ones not used as inline citations. However, it excludes any domains you’ve added to the ban list in Deep Research.
• Added an option to display all references used to generate Deep Research knowledge in each subheading. The list is only displayed in the text file and not uploaded to WordPress or displayed in the HTML file. So it’s ideal for fact checking articles.
• Increased the text limit for the Penny Arcade input box to 140,000 characters. I don’t know if that will break things. I guess we’ll find out!
• Added an option to disable image compression in the Bulk Writer, Penny Arcade, SEO Writer, and Product Roundup. Just select an AI image model and press the O button. Note: the setting is reset when you change a model, but your setting can be saved to a profile.
• Amazon has been getting trickier to scrape. So I enabled premium proxies to “pulling” the Amazon URLs for product roundups. If you are getting missed products or problems when scraping actual Amazon URLs then you might want to add Amazon.com to the “manually added domains” in the ZimmWriter ScrapeOwl menu and tick the PP box for premium proxies also. But I don’t think we’re at the point yet where we need to force that setting on everyone.
• The last patch fixed a JSON error occurring on non-English translations. But that fix broke Amazon product roundups using short codes. sigh. So now I’ve fixed it and it should work.
• Fixed a bug with loading a saved profile where it might throw an error when loading a custom outline that has multiple custom backgrounds in it. I guess it’s been in there for awhile and no one reported it (deju vu) so now it’s fixed.
• Lots of random bug fixes.

Released 2025-01-25

• Fixed a bug in the 1-click that was causing a crash due to the recent update. As a bonus, I’ve given the 1-click writer an upgrade and now it will auto nuke those pesky AI words and also bold your content.
• ZW has listicle detection in the Bulk Writer and Penny Arcade which override your H2 settings. But I decided to revoke this behavior for 0 and 1 H2. Now you will no longer get a listicle if you specify 0 or 1 H2 and your title is a listicle.
• Fixed a JSON bug in non-English translations.

Released 2025-01-24

• The Edit Custom Prompt (ECP) button never worked in the social media scheduler. No one told me though. So… now it works.
• Fixed a JSON error that could occur when using non-english and characters were located inside brackets [ ]. I previously did not encode this content in the event you were using a shortcode, but that caused issues with non-english. So now it’s encoded to an html entity.
• Added an extra backdated year to WordPress social scheduling, so now you can schedule two years back in time.
• Overhauled the professional mode style of writing.
• Overhauled the overall blog post output quality in many aspects.
• Overhauled outline generation.
• Dramatically improved scraping summarization quality.
• In the Bulk Writer and Penny Arcade, I added a 0 H2 option which will write a block of text with no H2 subheading title present. The length setting for both 0 and 1 H2 have been changed so that on a powerful model: tiny=279 words, short=340, medium=403, and long=465. However, each link (whether via a LinkPack or Deep Research) adds a new sentence, which is about 15 more words. So keep that in mind if you’re adding links. Also, the ZERO H2 works with Master Raw Prompts!
• Added a “None” option for the Intro in the Bulk Writer and Penny Arcade.
• Added Sonar Pro Online and Sonar Online for Perplexity API and removed their LLAMA3 models. Sonar Pro can be 2x or 3x as expensive as the normal Sonar model. The normal Sonar is the same price as before. As for OpenRouter, the Sonar models are not yet available on that platform.
• The slug generator has been overhauled. It now uses AI to determine the slug. In addition, slugs generated by ZimmWriter will now use AI to determine them (unless you provide them via the {slug=}). The previous approach was simply dropping invalid characters and replacing spaces with a dash. But that would end up with crazy long slugs.
• Patched a bug causing incorrect H# and a lone # when your article had identical subheading titles.
• Patched a bug causing a (#) after some links.

Released 2025-01-12

• Remove GPT 3.5 as it’s depreciated.
• Fixed a bug with some links that had _ being detected as bold instead of _.
• Setting an upper limit of H2 to 5 and the number of H2 to automatic will allow the lower limit behind the scenes to be 3. In all other situations, the lower limit is 5 (as it has been in the past).
• Added an option to exclude category ids that you don’t want used for the auto category determination. You can find the option in the WordPress menu.
• Removed ideogram 1 and added ideogram 2a and 2a turbo.
• Added romania to the SERP scraping dropdown.
• Added a new {profile=} feature to the Bulk Writer. Also it “loads” the profile so it continues… Also, there are a million moving parts and my code is getting nasty. So be on the lookout for bugs (settings not loading) when using this feature. If you find some, please PM me.
• Added a Bulk Slug Generator in the Vault. The main (and arguably only) use is when you want to use the pipe “|” trick in the Link Toolbox to link to articles not yet written, and you don’t know the slug. You can feed in your keywords or your titles to get your slug and build the url.
• Fixed a capitalization bug that woud not retain capital abbreviations for titles when the abbreviation was enclosed in parenthesis.
• Boosted the character limit for the Link Toolbox input box.
• Fixed a bug that would cause an unbolded extra ** in a link.
• Patched some odd prompt leaks that were happening with link insertions on stupid models like llama 70b.
• Fixed a bug where ZimmWriter would close on a lower end computer when it encountered a rare long HTML string from a page it was trying to scrape.
• Added a “today” option to WordPress scheduling.
• Improved markdown italic detection.
• You can now designate in the Bulk Writer the AI text model you want ZimmWriter to use for creating your image prompts. It is very useful when writing about something the AI doesn’t know about and you might want to user perplexity to generate your image prompt.

Released 2024-12-18

• Fixed a bug causing manually provided slugs to break.
• Fixed a bug that was causing ZimmWriter to close upon selecting a WordPress user in certain circumstances.

Released 2024-12-17

• ZimmWriter now has a WordPress plugin to help decode HTML entity encoded characters. You can download it here.
• Added an auto tag feature when uploading to WordPress. ZimmWriter will automatically generate 3 tags for the new post. You can disable this functionality in the WordPress menu in ZimmWriter.
• Added an auto category in the wordpress dropdown. When selected, ZW will put each article in one of your root categories (cap is 250 root cats) based on AI determination. Unlike the auto-tags, which creates your tags for you, the auto-category does not.
• Added auto {category=1,5,20,29} which allows you to specify multiple IDs (root cats or sub cats) and ZW will use AI to pick one to put the article.
• Fixed HTML tables from being broken at the last row.
• Bug fix that was causing ZW to force close when certain data was returned from WordPress (e.g., an inability to fetch the users). Also had the added benefit of helping me find some unoptimized code when connecting to WordPress.
• The Social Media Scheduler will now create hour, minute, second scheduling instead of simply relying on the post publish timecode (which could cause problems when multiple posts were published on the same day).
• Fixed a bug in the Link Toolbox that would cause weirdness when adding domains to the do-follow box.
• Amazon changed their layout so the product roundup didn’t pull the titles when pulling the 15 products. I believe I’ve fixed it but keep me posted if you run into any issues.
• ZimmWriter will now process markdown italics (underscores on each side) and convert them to the appropriate italic format when uploading to WordPress or generating your HTML file.
• ZimmWriter now features non-English slugs for WordPress.
• A lot for non-English WordPress compatability was done behind the scenes. I also created a WordPress plugin which will decode various HTML entities in the title, tags, and body content.
• Skinny Paragraphs would happen before Bolding, but it seems that using some models, applying Bold would cause the AI to eliminate the work of Skinny Paragraphs. So now I’ve flipped them. Bold applies first, and then skinny second. The insertion of links (Deep Research and Link Packs) still apply afterwards like usual.

Released 2024-12-06

• Deep research no longer embeds YT videos as that was nuking the paragraph that they were injected in. So now they appear as normal inline links.
• Fixed a longstanding bug that would prevent dollar signs from appearing in the subheading titles and faq titles.
• Found a memory leak bug that would occur when scraping hundreds of URLs and lead to eventual failure.
• Added more AI leak detection regarding link packs and deep research.

Released 2024-12-01

• So I’ve had pneumonia for the past month and my brain isn’t working as it should. The bolding worked in the HTML files but was stripped out on upload to your WordPress sites. I fixed that and now I’m going back to bed.

Released 2024-12-01

• Fixed some more prompt leakage from Deep Research.
• Fixed a bug that had been bugging me for awhile. The links by Link Packs and Deep Research would occasionally insert hallucinated URLs in rare situations. I thought I squashed it in the last update but I was wrong. Now I’ve stepped on the bug and it’s definitely dead for sure.
• Now you can bold and also assign a CSS class to your Link Pack links

Released 2024-11-29

• In the Link Toolbox, the temp disable ScrapeOwl box now will prevent you from getting the “blocked domain” when checked. The URL still might not scrape, or it might, but at least you won’t have to go into the Scrape Owl menu and remove the blocked domain.
• Fixed weird (https://yourlinkhere) that could sometimes appear with Deep Research or Link Packs on the output.
• Improved reliability and speed of both Deep Research and Link Pack insertions (but not Deep Research scraping; I can’t speed that up).

Released 2024-11-27

• Added more logic to OpenRouter to prevent article generation errors or failures when ZimmWriter encounters an error returned by OpenRouter.
• The Bulk Image Generator now doesn’t nuke numbers at the start of a title.
• The Deep Research tool now lets you specify up to 200 domains that you don’t want used as citations. The list of domains will be saved if you save it to a profile.

Released 2024-11-22

• Fixed the Bulk Image Generator and Text Discombobulator so they don’t resize the images anymore when compression is not selected.
• Improved image generation reliability with Fal provider to avoid server issues people are having.

Released 2024-11-22

• Added detection to determine whether your ScrapeOwl API is expired, out of credits, or having problems. It should now flag that at the start of a job, so at least you’ll know and can make a choice what you’d like to do.
• Fixed a bug where pressing the O button on a [NA] image caused a crash.
• Removed Haiku 3.5 from Ancillary choices as it was not intended to be there since it’s much more expensive than Haiku 3.
• Fixed a bug with Haiku 3.5 from OpenRouter that was showing 0 cost.

Released 2024-11-22

• Fixed a bug that prevented custom image prompts for subheadings not to work.
• Set the Deep Research Max links per article to 50 now, instead of the default 10.
• Fixed a bug that would cause the progress indicator to vanish in certain circumstances during blog. writing with Deep Research.
• Fixed {normal} to now work with additional variables, such as {save_text}.

Released 2024-11-22

• Added profiles to the Bulk Writer so you can save your settings and instant reload.
• Added the new and incredible Deep Research mode to the Bulk Writer which fetches real-time factual data and citations.
• Magic commands get a big boost and they now work with Unicode, such as emojis and non-English input.
• All image models in the Bulk Image Generator and Text Discombobulator now output to JPG.
• The compression button in the Bulk Image Generator and Text Discombobulator now works with all aspect ratios.
• Added an option to the Bulk Image Generator and Text Discombobulator to skip creating a prompt from your prompt. Instead, when you check the option, it will use your raw prompt (with placeholders still replaced) to create the images.
• The number of H2s in the Bulk Writer gets boosted to 50! Just be careful with this as it would suck if you’re writing using an expensive model and expensive images and something fails on the upload to your site. At least you’ll have the txt file (and possibly HTML) and jpg images which you can manually upload.
• You can now add your own models to OpenRouter. Just go to https://openrouter.ai/models and when you find a model you like, press the “copy” button to copy the model name – e.g., “openai/gpt-4-32k” and then paste it into the user choice model box in the Text API settings of ZimmWriter.
• Further improved the the prompt leak detection for OpenRouter models.
• Adjusted how scraping URLs for subheadings works in the Bulk SEO CSV. Now if a scrape fails on a particular subheading, the article will still write but without the scraped data for that particular subheading. Before, the entire job would abort.
• A new feature to the Bulk Writer is a custom job directory. It lets you direct the entire job’s output into a separate folder within your output directory. It’s ideal for organizing work, such as creating distinct folders for each site or client.
• Dramatically improved translation from non-English into English (e.g., when scraping a URL) or translating a title.
• Non-English slugs have gotten a big boost. Now if the language is of the kind that has English characters (while not being in English), the slug will remain in non-English with any accented characters now un-accented.
• Boosted the number of links that can be inserted in an article via a Link Pack from 10 to now 50.
• Boosted the number of links that can be inserted in an article subheading via Link Pack from 1 to now up to 5.
• Links (from both Link Packs and Deep Research) now insert on subheadings that have tables.
• Added a {fileid=123} variable to bulk writer title. Just add this onto the end of a title and all the files associated with the title will get prepended with that numeric id followed by a double underscore. So “How to Wash a Dog{fileid=1} becomes “1__how_to_wash_a_dog.txt”
• Added a ECP button to Text Discombobulator, SERP Discombobulator, and Social Media Scheduler. The ECP button allows you to edit your custom prompt in those menus, without having to switch back to the Bulk Writer.
• Added Google location options to the SERP Discombobulator.
• Added the ability to call a custom image prompt on a specific subheading. You will have to name your custom image prompt in the Subheading database using the “P” button in the Bulk Writer in a {img_bla_bla} format and then use {img_bla_bla} at the end of a subheading in a custom outline. Note that at the moment it is not possible to associate custom image models and settings with your custom image prompts. So the subheading prompt you select in the dropdown of the Bulk Writer will control all subheading image prompts.
• Updated the default ZimmWriter image prompt for featured images and subheading images.
• Other random bug fixes.

Released 2024-11-12

• Fixed a bug with non-English where image filenames would not be created on upload.
• Fixed a bug with non-English where sometimes the slug for the article would get messed up.
• Fixed a long-standing bug with handling certain Unicode characters which might have resulted in your blog posts displaying ? characters. It might also have a huge affect on translation quality to non-English too.

Released 2024-11-09

• Patched a bug with master raw prompts when combined with additional outline elements.

Released 2024-11-08

• Patched a small bug with Flux Ultra’s RAW mode not applying in 774 for Fal.

Released 2024-11-08

• Added the amazing Recraft AI image model using Recraft’s own API and also Fal’s and Replicate’s APIs.
• Added Flux Ultra to Fal and Replicate.
• Added Anthropic’s Haiku 3.5 model (which is more expensive than the 3 model).
• Added SEO keyword calculation on SERP scraping for: Australia, Brazil, Canada, Haiti, Netherlands, Norway, New Zealand, UAE, United Kingdom
• Now ZimmWriter will display all AI text and image models, but prepend [NA] before ones you do not have access to since you may not have added an API key. A few long time ZimmWriter users didn’t even know that we have all been using Anthropic’s Claude for over a year because it was simply not displayed in the dropdown since they never checked the API menu (and didn’t read the changelog).

Released 2024-10-12

• Bugfix to SD 3.5 when using (S).

Released 2024-10-12

• Added Stable Diffusion 3.5 Large to (S), (F), and (R) providers.

Released 2024-10-17

• Fixed an inaccurate guestimate of how much SD costs.
• Added functionality to better detect manually added images in a subheading for the social media scheduler.
• Solves shortcodes from not rendering (e.g., AAWP or AFFI)

Released 2024-10-14

• Fixed a crash that was happening when ZimmWriter encountered a YouTube video whether in the Penny Arcade or via SERP scraping.

Released 2024-10-13

• Updated YouTube transcript scraping to work better.
• Added the Grok 2 model via OpenRouter. However right now it’s rate limited for some odd reason. Maybe that will clear up in a few days. It’s an interesting model. Have fun with it.

Released 2024-10-10

• Fixed a crash that would occur when using the Text Discombobulator with a the skip line @ trick.

Released 2024-10-09

• Added {save_text} to the custom outline and {load_text} to custom image prompts.
• Fixed a bug that would prevent listicle detection for a master {raw_ prompt or {normal} prompt.
• Flux 1.0 Pro updated to Flux 1.1 Pro which is cheaper.
• Added image generation to the text-discombobulator.
• Updated GPT-4o to reflect the lower cost.

Released 2024-09-29

• Fixed a bug in the yourjobname_complete.csv that the social media scheduler would generate. It would format the social share date weird, which would cause problems importing into tools such as Publer.

Released 2024-09-27

• Users were getting too many links inserted when using a linkpack added to the title and/or subheadings. Now I have fixed this so it will obey your link pack menu or default to 5 links.
• Added a Social Media Scheduler

Released 2024-09-16

• The 10.762 update broke master raw prompts and this patch fixes them so they now work again.

Released 2024-09-16

• Fixed some abbreviation capitalization bugs in the subheadings and FAQ
• Fixed a big bug with link packs where links ending with / were re-scraped when updating a pack.

Released 2024-09-14

• Lots of little fixes, prompt improvements, etc.
• Added a {normal} tag for the Custom Outline. A video will follow in a few days.

Released 2024-09-02

• Patched an pop-up error message that would occur when you tried to add a new WordPress site.

Released 2024-09-01

• Patched a major bug with creating link packs with a I where your summary might have had a comma. If it did, then things got screwed up. You’ll have to remake that pack.

Released 2024-09-01

• Fixed a UI bug that made the disable meta description in the WordPress upload section not visible.
• Added some logic fixes when enter/deleting/modifying saved WordPress sites.

Released 2024-08-31

• Added a superpower to the Link Toolbox. When you build a link pack, what if you want to add links that can’t be scraped or links that don’t exist yet (e.g., the article isn’t published)? Well now you can accomplish this using the pipe symbol and a summary. Simply add a | after a URL along with a single line summary of the URL. It doesn’t have to be a long summary, strive for 50 words or less. When you start the processing of the Link Pack, any links with the pipe and a summary not be scraped and the summary you provided will be used by ZimmWriter.
• Adjusted the HTML code ZimmWriter generates when uploading to WordPress (and also outputting HTML) when you have selected non-16:9 images. Now if you select a 9:16 image for example, it should appear correctly inside of WordPress.
• Added the ability to add multiple users to a single WordPress site.
• Added the ability for any WordPress “admin” user to publish as any other admin, editor, or author, even if you have not added them to ZimmWriter.
• Added the ability to designate who you want to publish the article as by adding {author=the-users-wordpress-username} after a title in the bulk writer. The username must exist on WordPress, otherwise the default user you are connected as will be the author.
• Added two new databases to store image prompts. You can access them using the P button adjacent to an AI Image Model dropdown. In the Bulk Image Generator, the “database B” corresponds to the featured image database and “database C” corresponds to the subheading image database. Using either of these will override anything in the “Set AI Image Prompt” menu.
• Added the ability to input non-english titles into the bulk writer. But there are some unfixable bugs on my end that I simply cannot solve until ZimmWriter is recoded in a new language. The biggest bug is that in some languages (e.g., Hindi) the title is not parsed correctly by ZimmWriter because the Regular Expression library is outdated and I have no way to fix this at the moment. Note: everything else you input in ZimmWriter should be in English.
• Added the ability to select the AI model you want ZimmWriter to use for translating the output. ZimmWriter would previously use your ancillary model, which isn’t as smart as some of the more powerful models such as Claude Sonnet 3.5.
• Added some logic to prevent ZimmWriter from citing the article currently being written in the event you are using a link pack with that future article inside. Works great on english titles and titles using {slug=}, otherwise it will not work.
• Added a “Temp Disable ScrapeOwl” option in the Link Toolbox. Sometimes disabling ScrapeOwl can avoid blocked scraper problems. But as always, make sure the domain isn’t already blocked in the ScrapeOwl menu.
• Added some additional guidance intervals to some of the AI image models which could accept it.
• Added a “seed” option to the AI image models which can accept it.
• Bug fix to all AI image model options where the last option in a dropdown would not apply even if selected.
• Bug fix to the txt file generation so that links are in proper markdown format and should convert properly.
• Bug fixes to the URL processing in handling encoded characters inside of links.
• Bug fix so ZimmWriter can now display up to 1,000 WP categories (but it will be slower). Previously you were limited to 100 categories.
• And last but not least… boosted H2 limit in the bulk writer to 25.

Released 2024-08-26

• YouTube transcript scraping should work again using Scrape Owl. It now uses premium proxies so each scrape will take about 15 credits.
• Added Ideogram API. Note that when you add Ideogram, they will pre-bill you $40 for credits. You can obviously use those credits, but just realize that you will be charged $40 up front.

Released 2024-08-21

• Fixed a bug that was introduced in 10.750 which failed to remove variables attached to a custom outline subheadings during the backend prompting.

Released 2024-08-20

• Fixed a crash that would occur when using non-linkpack links in the penny arcade.
• Adjusted the first person tone of voices a bit.
• Fixed a bug that would cause ending punctuation on a paragraph to vanish when skinny paragraphs is off + a link is inserted.
• Patched some translation leakage that could occur.
• Adjusted bolding for list items.

Released 2024-08-19

• While we’re at it, adding the image ratios in 10.751 screwed up how Scraping Surgeon images were treated (e.g., Amazon images). Now that’s fixed. I wish I caught it ten minutes prior to pushing out 10.752 but such is life.

Released 2024-08-19

• Emergency patch. The prior version had a bug in the download file function which would download missing files such as the database driver in the event the file was deleted.
• But as a result of the emergency patch, everyone gets these treats:
  • Perplexity 405b Online
    • Note: the naming of the perplexity models changed, so if you had any custom outlines with the models listed in-line, then you should update them accordingly.
  • Hermes 405b

Released 2024-08-18

• Adjusted the ‘random’ setting for Amazon reviews to now choose between 3, 5, and 10.
• Fixed a bug on Wine (e.g., Whisky) where you couldn’t input 2 digits (e.g., 15) in the subheading box for H3, H4.
• Added a patch to better avoid saying “tested” in the title when writing a product roundup.
• Tweaked link-pack insertion to improve relevance detection in both linknado and non-linknado mode.
• Fixed a bug that prevented link packs from being applied in custom outlines at a subheading level – e.g., {lp_linkpackname}
• Added the ability to now specify any supported aspect ratio for both featured images and subheading images in your blog posts, depending on your AI model. Note that if the images end up “appearing” squished, that’s a theme problem, not a ZW problem as they should upload in the correct aspect ratio.
• Added an option in the Link Pack to toggle whether you want the links to open in a new tab or not.
• Added a “MASTER” raw prompt mode for ultimate power.
• Fixed a bug that would prevent bold words from applying in certain circumstances.
• Added a new “prepend” variable to custom outlines.
• ZimmWriter can now scrape URLs that have non-English letters inside of the URLs, which means you can now scrape them and also use them in Link Packs! I am surprised I was able to get it to work. I’ve tested a lot of things and it doesn’t appear this caused any other problems down the road with scraping links in general. Let’s cross those fingers!

Released 2024-08-13

• Added a secret button for advanced users in the Style Mimic menu who like to live dangerously. If you know, you know.
• Added the new Flux AI image models provided by Replicate and Fal. Note that the Flux Dev and Flux Schnell are uncensored.
• Added Link Pack technology. Create a virtually unlimited number of link packs each with up to 1,000 links and cause ZimmWriter to automatically inject the links into the articles as it writes. Please read the link pack directions first. Make sure to occasionally back up your link pack directory (e.g., close ZW, and just backup the folder somewhere).
• Added Perplexity 3.1 LLAMA models.
• Long overdue patch to allow quotations in the output.
• Many tweaks and patches to improve quality of overall output.
• Many patches made to LLAMA processing for prompt leakage.
• Fixed a bug that would prevent an upper limit of 5 subheadings + automatic mode to function properly in the Bulk Writer.

Released 2024-07-25

• Patched a hard to find bug that was causing the switch ancillary model window to appear when scraping a site in non-English script (e.g., Hindi).
• Removed the LLaMA 3.1 405b model from Groq since it’s not working yet even though it’s on their site. It was just leading to confusion.

Released 2024-07-23

• Added LLAMA 3.1 70b and 405b to both OpenRouter and Groq (Note: the 405b is unavailable on Groq for free plans. I anticipate in a few days they will add paid plans, but when they do, the cost will still show zero. I am not sure how to account for this at the moment.)

Released 2024-07-20

• Fixed a bug that did not display the ScrapeOwl menu in the settings.
• Fixed a bug in custom outlines where you had multiple subheadings one after another, each with a raw prompt, and each raw prompt was designated as not to upload to WordPress. In that situation, only odd raw prompts (not even) ones would be skipped from uploading to WordPress. Yeah, I dunno. It’s too hard to explain. Basically, I fixed some stuff and now it works.
• Further improved determining whether to convert numeric items in the output to HTML lists or not.

Released 2024-07-19

• Unbundled OpenAI from ZimmWriter (very hard to do) so that you no longer need an API key from them if you don’t want. Use any API platform you want. This is great for people in countries like China who OpenAI doesn’t allow to get API keys.
• Bumped the custom prompt limit database storage up from 50 to 500.
• Patched the conclusion which could occasionally get too long.
• Added a background waterfall. In the SEO writer and also in the newly added Custom Background, if you gave an H2 a background, but not a lower level H3 or H4 a background, then the H2 background would not be feed into the children H3, H4. So if you were talking about something the AI didn’t know about, the AI might be clueless. I have changed this behavior by adding a waterfall. Now, ZimmWriter feeds the higher level background into the lower level background when the lower level background is empty.
• Added a raw prompt feature with prompt stacking, to the custom outlines.
• Added a custom background and custom global background feature to the bulk writer.
• Added an area in the ZimmWriter options menu to rename the FAQ and Conclusion to whatever you want. Some people wanted this to avoid a Google footprint.
• Added a link on 1-click, bulk, and penny arcade on completion to open the folder where the articles were saved. Just click the link to open the folder.
• Added the ability in a custom outline to designate a specific model to a apply for a specific subheading. Also added a copy model button to the Text API menu in the ZimmWriter options menu which copies all of the models to the clipboard in {model=} format for easy use in a custom outline.
• Added the Stable Diffusion Ultra model.
• When converting to HTML and also when uploading to WordPress, some nested lists format better to avoid the 1, 1, 1, 1, that sometimes appears.
• Added GPT-4o Mini to OpenAI and also OpenRouter.
• Reworked the secure mode which is useful for letting a VA use a ZimmWriter license.

Released 2024-06-28

• The only model with an additional cost per 1,000 requests is Perplexity Online (OR) at $5 per 1,000. So I’m now adding the fractional amount onto the total cost summary you see in the text file based on the number of calls to Perplixty Online through OpenRouter for the article.
• Fixed a GUI typo that incorrectly stated the number of articles being written in the pop-up when URL Merge is being used.
• Patched the Nuke AI Word function as it was performing subpar on GPT 3.5 and some other models.
• Removed the OpenAI mode (trial/paid) status because it’s no longer needed.

Released 2024-06-27

• Added OpenRouter.ai (OR) along with a hand-selected group of models accessible through OpenRouter.
  • What is it? It’s a passthrough to different AI model providers. So imagine you are in a country that does not allow access to OpenAI. Using OpenRouter, you can now use OpenAI’s models through OpenRouter. The best part is that you only need an OpenRouter key and not multiple keys for each provider on OpenRouter.
  • As for pricing, the model prices can change on OpenRouter. So it’s best to check their website. Each time you restart ZimmWriter and load a model, you’ll see a pop-up telling you that ZimmWriter is loading the model, and it will also fetch the most recent price. The price will be reflected in the txt file. If you want realtime prices, please check the OpenRouter website.
  • The models through OpenRouter are subject to change at any time.
  • Two words: Perpexity Online!
• Added a URL Merge feature to the Bulk Writer. Imagine the Penny Arcade, but now you can merge multiple URLs into one article (kind of like SERP scraping, but with user provided URLs).
• Fixed a very hard to find bug that dealt with SERP scraping when you might be running multiple licenses of ZimmWriter on separate computers simultaneously and scraping using ScrapOwl but with only a one concurrent ScrapOwl session plan. Phew.
• Moved Custom Outline storage from settings.ini to the database. Now you can store 1,000 custom outlines if you want.

Released 2024-06-20

• Added Anthropic’s Sonnet v3.5
• Fixed a glitch in the advanced triggers that wouldn’t re-display your current voice settings even though it was still applied on the backend.
• Fixed some leaks for advanced triggers and magic commands
• Added Taiwan to SERP scraping.
• Tweaked the error message when an upload to WP failed, so that we now exit and see what wrote and what did not write. Or at least that’s the plan.
• The clear all data button now does a much better job of clearing all data.

Released 2024-05-21

• Fixed a glitch that would cause a trailing ” on a title in TAG
• LLAMA was outputting lists using real bullet points which threw off ZW’s code when it came to identifying them on conversion to markdown or uploading to WP. Now that’s fixed.

Released 2024-05-17

• Fixed a bug that was causing GPT-3.5 to still be used (and not LLAMA3 or Haiku) for the ancillary task of calculating keywords.
• Fixed a bug that was causing a loaded save slot in the Bulk Image Generator not to populate.

Released 2024-05-13

• Added GPT-4o

Released 2024-05-12

• Patched Groq to better give pros/cons in product roundups.
• Fixed aspect ratio not working in bulk images for SD model (worked in SDC, SDT, and SD3)
• MAJOR improvement to LLAMA and Claude models in all aspects. Just be aware that since these are non-OpenAI models, you could still get the occasional glitch since when they interact with ZimmWriter prompts. Also, when using these models you may have to re-work your custom prompts. So YMMV.

Released 2024-05-10

• Emergency patch to fix a bug that would prevent entering a new OpenAI key.

Released 2024-05-10

• Added Groq and Anthropic AI models. Here is the guide to get started.
• Tweaked H3/H4 to more closely follow section length settings.
• Bumped up the # of text tags in Scraping Surgeon from 5 to 15.
• Fixed a bug that would only cause one “con” for a product roundup or trigger product layout.
• Fixed a bug with subheading images that would not obey the “every other h2” setting when h3/4 were used.
• Fixed the scraping of YT video transcripts that Google broke when they changed the YouTube code.
• Added better handling of rate limits under OpenAI such that if you hit one, you can wait out the 24hr wait period if you’d like without stopping ZimmWriter.
• Added aspect ratios for stability models in the bulk image generator in the AI vault.

Released 2024-04-27

• Added a Bulk Image Generator feature to the AI Vault.
• Bug fixes.
• Added SD3 (Stable Diffusion 3) and SD3 Turbo, but mind the pricing. You need to test to see if it’s worth it in your niche.
• Added the ability to copy data in from the clipboard for the Local SEO Buffet! You can see a guide to the required formatting here in the Local SEO Buffet guide.

Released 2024-04-16

• Patch to accept new OpenAI API keys that went into affect in the past day or so. The backstory is that OpenAI has changed the length of API keys because they now have “organization” keys. At the moment, ZimmWriter doesn’t support adding an organization. But the keys now will work at least.

Released 2024-04-15

• Emergency patch because the new Stable Diffusion model costs 3 credits (10x) more than the prior Stable Diffusion model. So I brought the old model back as an option for you to choose. In the dropdown “SD” is the original model which costs .3 credits (or $0.003) per image and “SDC” is the new Stable Diffusion Core model which costs 3 credits (or $0.03) per image.

Released 2024-04-15

• Updated the OpenAI models so they are all pointing to the most recent versions.
• Updated Stability to include the new “Core” model which is amazing.
• Updated the Local SEO Buffet to better attempt to state that the business services a location instead of saying its located in the location.

Released 2024-04-04

• Fixed a bug in Nuke AI Words that wouldn’t allow apostrophe in the word list (so you might want to check you list and update it with words that are supposed to have them).
• Fixed a bug with headings that wouldn’t allow apostrophe in some situations

Released 2024-04-04

• Added a “lite” mode checkbox for SERP scraping. It’s for people with slower computers that take too long for SEO keyword generation. It will use 2-3 word n-grams instead of 1-4 word n-grams, which should speed things up.
• Added an intro selection feature to choose either the standard intro (which was what ZimmWriter had before version 10) and a search intent focused intro which is what was present in version 10 and attempts to answer the search intent as soon as possible.
• Improved the standard intro.
• Improved the Nuke AI Words feature.
• Reduced the chance of longer FAQ questions.

Released 2024-03-19

• Fixed issue where bolded markdown words with asterisk were not changed to appropriate HTML tags in the HTML file when using HTML output.
• Even more improvements to bold quality.
• Removed bold from sections with tables.
• Fixed a bug that would allow selection of auto style or write in the style of when using mimic in a certian situation.
• Fixed some uppercase issues with certain words in titles and subheadings. Progress people, progress.
• Enhanced performance of Nuke AI Words when using it with Style Mimic.
• Bold AI words now works with product roundups.
• Dialed down intro length a bit.

Released 2024-03-16

• General patches.
• Search intent update to intro and key takeaways to help searchers.
• Search intent update to add bolded text to help searchers.
• Search intent update to AI generated SERP outline to help searchers.
• Created Style Mimic. Read the directions. YMMV.
• Added Nuke AI Words to all keyboard triggers (see the advanced trigger menu for the checkbox).
• Added Nuke AI Words to the Local SEO Buffet.
• Adjusted naming for some CSS class wrappers.
• Images now attach to the post when uploading to WordPress.
• Added to Amazon: amazon.sa
• Added to SERP: Saudia Arabia, Singapore, South Africa, Spain, Serbia
• Added the ability to start a line in the discombobulators with a @ which will act as a placeholder and skip that line from AI processing.
• Added option to output a HTML file in addition to the text file (check the options menu).
• Improved the quality of SEO Keywords to Blog Post Titles by adding a second pass. So it’s more expensive, but higher quality.
• Added the ability to rename “Key Takeaways” to something else of your choosing in the options menu.