Final Thoughts on Sudowrite

This will be the final post on incorporating Sudowrite into my writing. I’m not going to stop using it, but I do think I’ve gone in-depth enough to give folks an idea of what it’s like to start using a tool like this. I’ve also come to some final conclusions about how I’m getting use it that I don’t think will change. I went back to dictating/my process. While the first four chapters leaned very, very heavily on Sudowrite, as I got further into the book the harder it became. Starting an e-book comes with a certain amount of freedom, especially if the books are character-driven. For me, the specifics of the scenes were far less important than what those scenes showed regarding where the characters were in their journey with one another. So, I had more flexibility at the start of the book than I had in the middle. By chapter 6, I was no longer using Sudowrite to spin up scenes and had gone back to my original process outlined elsewhere in this blog For creation. The more I gave it, the better it was. Despite being pretty attached to expand, I learned to develop an appreciation for wormhole. The expand option in the software has very little to work with as far as my own voice whereas wormhole seems much better at attempting to mimic my voice as a writer. Description, too, seemed to work better the more text it had. Sudowrite was most useful for me when editing I tend to tell stories through dialogue—my books come in at around 60% dialog on average. I’ve made a conscious effort over the past several years to go back and flesh out descriptions, and my readers have seemed to really enjoy that. This is where Sudowrite is indispensable for me—I rarely use its description word for word, but it gives me descriptor ideas that make it well worth the money. I think there are places it can go. There are some things I wish Sudowrite had, and I’m sure the developers might add them in the future (or another tool more specialized to world-building or grammar might come along utilizing GPT-3). That includes character generation, plot beat generation, mystery seed generation, rephrasing suggestions (like a sentence or paragraph thesaurus), and intelligent descriptor expansions. I know some authors with far more technical know-how than I are already creating their own tools for their own specific needs. I’m just, unfortunately, not that smart. 🙂 Don’t get me wrong. I think Sudowrite is a complete tool with a robust focus, and I definitely recommend it. But having played around with this tool for more than a month, I feel like I just scratched the surface of what it could do and how it could improve my writing. I will keep using it. Just not for generation, I think. Editing, adding, adjusting my writing, yes. Generating? Maybe not quite yet.

Finding the AI Balance & Keeping Organic Connections

So, this is probably going to be one of those frou-frou writer observations. The last chapter I wrote utilizing Sudowrite heavily was Chapter 4. As I’ve mentioned before, in starting my latest book I was incorporating Sudowrite more and more and more. I think with chapter 4, I reached beyond a line I was comfortable with. I noticed as I was writing and editing, I was struggling. I didn’t emotionally feel the same connection with the characters in the storyline of this book as I had in my previous books. I write one book at a time and one series at a time because I really do inhabit, mentally, that world. Falling asleep at night, I think about the next phase in the next chapter. I sometimes dream about the different directions the book will go. In short, any quiet time I have you can pretty much bet I’m going over the book in my mind. Chapter 4 was where that connection faded pretty dramatically. I no longer felt like I was writing the book. I look back over it, and I see that the beats are mine and the things that happened are my ideas, but some of the things I incorporated based on the AI’s suggestion I just didn’t feel a connection to. I wrote chapter 5 today utilizing Sudowrite as a contributory tool, but I didn’t use wormhole to come up with directions in the scene. I hewed pretty closely to leaning entirely on “expand” (Again, expand is the tool that gives me a scene based on what I tell it to write, or from what I’ve already written. That last part of actually an “off-label” use, I guess. The directions say to give it a summary. A lot of the times I give it a hundred words of text I’ve written and let it go, and a good portion of the time it comes up with something really good.) Anyway, I felt that connection to the material re-forged. I’m not going to stop using Sudowrite, but I do think I found the hard limit to how much I can incorporated suggestions without losing the thread of what I’m trying to say in the book or losing that feeling of connection and ownership. While “wormhole” is ready-ish for prime time and “expand” is an experimental lab that could disappear at any time, I can say without a doubt that for the way I write, I find expand much more useful. That could be, though, because of the type of writer I am. I personally rarely run into plot blocks, and I generally don’t get stuck on the story itself. Where Sudowrite helps me is coming up with actual words that present things in a slightly different way than I normally write them to help make the words themselves not boring. I won’t kid you. It was a disconcerting experience feeling disconnected from my own book. It was temporary, but it made me grouchy for a couple of days …

Continue reading…

Has Sudowrite helped my production? Yes.

I have a rigidly disciplined writing schedule. I make a not inconsequential amount of my income from preorders, and I try and make sure I have the next three books up for preorder at any given time to allow people to grab them. Two of those books are always planned, but not written—hence, rigid discipline is needed to get them out when I claim they will be out. My spreadsheets rule my life. I have a defined amount of time to write and self-edit the book—49 days. A defined amount of time in between books as a grace period in case I need it—7 days. 42% of the 49 days are “creation” days, which means in a single week, I should create at least 3 chapters. Any less, and I’m behind. Yes, there’s another spreadsheet for that, too. Like I said. My life. Ruled by spreadsheets. How It All Was I’ve always wanted to get faster, but when I tried, the writing suffered quite a bit. This schedule (work 6 days a week with three writing and three editing days) seemed to produce the best outcomes for the book and (stress/healthwise) for me. There are clear indicators in these spreadsheets all over to make sure I’m on schedule—the percentage left to do should not exceed the time left to do it on the second sheet. The “time left for XX chapters” and “written as of today” better add up to at least 20. And so on. As I mentioned before, my old method was: Dictate a full chapter, edit that chapter the following day, then dictate, then edit. I didn’t like stopping in the middle of a chapter, even if I was stuck. I don’t care whether I write 500 words an hour or 1500. My day’s job is getting that chapter written and done, or edited and done. If it took me an hour or six hours, that’s what it was going to take. The dictating is…problematic. Did you know introverts can get depleted talking to thin air? Apparently, it is an actual thing. I can do it, the dictation, but I dislike it greatly. (Carpal tunnel and De Quervain’s tenosynovitis make long typing sessions extremely painful and potentially damaging, hence the dictation.) I’m also incredibly self-conscious about anyone hearing me, so my husband was not allowed to get out of bed/wake up until I was done. We even have a multicolored nightlight. If it was red, he could go to the restroom but beyond that? Red meant everyone stayed quiet and stayed away. How It Has Been With Sudowrite So, I haven’t touched Dragon at all in the creation of Book 3. Not once. That’s been…pleasant. (My husband would also like me to mention the non-use of the nightlight has been nice.) The addition of Sudowrite has also changed the way I write quite a bit. Previous to Sudowrite, I needed: Total and complete silence. No music, no background television, no talking, no nothing. To begin and end …

Continue reading…

Book 3, 2.5 Chapters in, all AI-Augmented. My current process.

Previous to using Sudowrite, my process went something like this: Dictate chapter. While dictating, I might make use of MasterWriter, a personal database of books in DocFetcherPro, or various thesaurus and prose helper books if I got stuck. Come back the following day and edit chapter. This usually resulted in adding about a thousand words. My dictating tended to focus on minimal action and dialog to sketch out the full chapter beats. The following day I would add the nuance, humor, and descriptors, as well as tighten up the dialog. Post edited chapter to my beta readers and let them tear it apart. Come back in a few days and incorporate their suggestions or correct any typos they may have found. At the end, the whole book goes to an editor, natch. I am a very linear writer. I write one scene after another, one chapter after another, and I deal with each scene before going on to another one. It’s very rare that I go back into a finished scene or chapter to add anything other than red herrings or stronger indicators for the mystery if I think the reader may not have enough to play along and figure it out. This book, which is Magic’s a Hoot, is going something like this: I start most scenes by spinning up something from “Expand” in Sudowrite, which is experimental. (Expands a summary into a scene. This is useful if you know what happens in a scene, but are having trouble rendering concrete beats and details.) The AI is (as experimental features can be) hit or miss. I find the simpler the summary is, the better the AI tends to do. It does, though, get drunk and lose touch with reality at some point, and the whole thing turns into a word salad. At times, though, what it gives me is remarkably on point. I take from this what I can, and flesh it out. From here, I usually take small pieces and feed them back to sudowrite as I make it my own, and then run “Wormhole” on the smaller section. (Wormhole: Given a passage of text, Wormhole generates possibilities for what to write next.) Lather, rinse, and repeat step 2 until I get to the end of the chapter. I let the chapter rest a day and then come back and do my normal steps 2 through 4 with the exception that I may refer back to Sudowrite for “Describe”. (Highlight a word or phrase and Description will suggest a few ways to describe it. Description works best when highlighting a short phrase within a paragraph. But it can also work at the paragraph level if you want it to suggest descriptions for a broader set of objects in your scene.) I have given my first chapter to my beta readers. They did not seem to notice any difference in the writing style. They didn’t comment on anything being different, odd, clunky, weird, not funny anymore. Nada. I expected …

Continue reading…

Where the AI ethics began to get murky.

So, a few days ago I said this: Did I feel like a computer was writing my book? Not remotely. It wrote words, but they weren't MY words-my tone, my voice, my vocab, my cadence. It was like a menu, and I chose what I wanted. Then lobbed ketchup all over it to make it my own. — Leanne Leeds 👻 (@LeanneLLeeds) June 3, 2021 And I think I spoke too soon. I finished Owl’s Fair yesterday, and on a lark, I decided to play with Sudowrite. Could the AI, I wonder, actually write? What could it give me to start? Sudowrite has an experimental lab with the following directions. “Write a scene summary (less than 100 words) in the Editor, highlight it, and click the Expand button. A good summary succinctly describes the characters, setting, conflict, and how it resolves.” So I did. I delete the documents as I finish my scenes, but I provided it with something like “This scene is written in the first person from Astra’s perspective. Astra Arden, a witch, and Emma Sullivan, a police detective, are looking for a stolen necklace in a park. Astra finds it.” Or something equally as simple and benign. You can see a smidge of it below. I hit the button and let the AI do its thing. On the left is my initial cursory editing of the scene based on what the AI (right) gave me. It was far better than I suspected it would be, but it fell short of being able to literally map out a whole scene. So, I went at it a different way—I described the beats and exchanges in the scene and spun up sections. As I got what I wanted, I edited and placed edited sections into wormhole (which is not experimental and seems a bit less hit or miss) to spin up a bit more. Though halting at first, by the end of the afternoon, I had a full chapter of Magic’s a Hoot primarily utilizing this back and forth. The bones of it had been generated by the AI. Everything started with the AI. All of it. Like I said—it started out as an experiment. Just something to do on a Friday afternoon, just playing around with a new tool. And yet, reading it…it was a perfectly good chapter. I edited, shaded, changed…nothing seemed out of place. It’s like the AI provided an outline, a scaffolding to sculpt on, and I took it and carved and colored the nuance. But I will admit to you, I paused, stared, and felt vaguely unnerved by it. I never intended, when I got the tool, to use it in this manner. And yet I had, and I had a 3500-word head start on a book I needed to write. I’d also doubled my production—I write a chapter a day. Yesterday, I “wrote” two. (Ironically, my strength tends to be editing—it’s the part I find fun, where the colors and subtlety really get sussed …

Continue reading…