This guide covers going from nothing to an uploaded video clip for TikTok. Advanced techniques and full-motion video are not covered here; this is more basic. However, methods used here can be applied to achieve more. I may cover video generation in a future guide.
Decide Your Subject Matter
Sticking with something you know is best, although producing videos based on trends is not uncommon. You will be targeting ~60s for the clip duration, and the amount of content you can squeeze into that time frame depends on voice cadence. If you’ve read the previous Audio-Player guide, you will have some insight into how to clone your own voice or use a system like ElevenLabs. Or, simply record yourself. Script length needs to be adapted, regardless.
How AI can help
AI is useful in many places here. The first is, obviously, the production of audio if you’re cloning or using voice generation. It can also help with determining how much you can fit into 60s.
Tip: Tools exist that can help you find out what’s trending on various sites. However, many of these are paid. Free tools such as SerpApi allow you to do your own searches with the right configuration. Google Trends is a good place to start. Designing your own system and aggregator is very possible. AI tools can help here.
Tip: You may need to modify audio speed or otherwise edit your audio. An Audacity alternative like Tenacity, ocenaudio, or wavacity will work.
Create the Script and Audio
You will want to put in natural subject shifts and pauses to cover ground and to seed places for image transitions. This means perhaps not writing as you would normally in a forum post or somewhere similar. Rather, you are fighting against limited time and the algorithm. Choose your words carefully. Then use your chosen method to produce the audio.
How AI can help
AI can help a lot here. Write your whole script? No! What you want to do is give AI a lot of your content and then have it ask YOU questions about how you write. Have some samples made and correct it until you have a good voice.md. Letting the AI know your intent helps more, especially if you have little to no content. You can use pre-made AI, AI voice generation, or AI voice cloning to help if you don’t want to voice record.
Tip: You should already be making notes about where you would want transitions or image changes. The script flow should make sense for this. Also, if you intend to have a regular intro, you should account for that in the timeline. This intro should be made separately for ease.
Make Subtitles
Once you have the audio created, you may want to make subtitles. This is useful because it can help direct your image placement on the video timeline later. AutoSubs for DaVinci Resolve is a good way to do this. You do NOT have to use this to import into Resolve, and instead can just transcribe and make an .srt file. Direct import if done right will give more control with Text+.
How AI can help
AI will have access to your script and could generate subtitles here, if needed.
Tip: Start with the Small model for AutoSubs. It’s fairly accurate. If you need better accuracy, move up to something larger if your system allows. AutoSubs has many other options you can set, including multiple languages.
Create Images
If you made clean points for transitions, you now have to fill them in with images. You can use your own photos, draw things, use free or open-source stock photography, or generate. The images should match the subject at hand.
How AI can help
You can use AI to create the image prompts for you. In some cases, you may want to be mindful of dimensions and the aspect ratio, as TikTok is 1080×1920 in portrait. Gemini and Grok are reasonable image sources, and you may prompt these by hand if so desired. Your tools can also use API or, alternatively, an MCP setup like Playwright to further automate this workflow. If you prefer to generate yourself, you may end up using something like ComfyUI and models off of Hugging Face such as Juggernaut, DreamShaper, Flux, and more. Setting this up can be complex and does require that you have the right hardware. You can train models especially for your subject matter with LoRA, but that will be fully explored in a future guide.
You will also want to resize your images for the intended output, e.g. 1080×1920. AI can help with this if you don’t do it manually. This can be especially useful for cropping out Gemini watermarks in a predictable fashion with a script.
Assemble the Timeline
You can now run Resolve or the editor of your choice to start assembly. Resolve is a good free option. If you are familiar with video editors you should be fine; if not, you may have to study some videos to learn. Assuming you know your way around, it’s now a matter of dropping in your audio elements and then the images in a video track. Custom effects and text can be added as needed. Transitions are as simple as clicking element borders and picking the right one. You have to stretch each image’s timeline to match the script to be sensible, but in general this is straightforward.
How AI can help
This is where AI use gets tricky. You can in fact use Resolve’s API, called as a script with the Free version, to automatically populate the timeline. You can have your agent build this script. For images, you will want to convert them to video temporarily so they take up the right duration. The duration is determined by where images are swapped, so you will want AI to know these times from the context/text chunking. It can then get the frame count right based on your FPS. Some manual adjustment may be necessary.
Tip: Make sure the project format matches your desired output. For TikTok, this will be 1080×1920 at 30FPS for the highest fidelity.
Deliver the Final Format
Once you have everything the way you like it, and that includes adding the subtitle/.srt if you want that, you need to deliver in the right format. Usually this will be MP4, H.264, 30FPS. If you want the subtitles ready to go, set the subtitle format to “Burn into video.” Note that TikTok’s web editor does not have a caption/subtitle feature, so burning them in may be necessary; the mobile app does support adding captions after upload.
Tip: Instead of subtitles via .srt, you can have them set as Text+ which is more flexible. AutoSubs has some flexibility here.
Upload the Video
With free Resolve you can automate everything with a few clicks, minus text placement. With the Studio version you can potentially automate further. You can even have code or an agent handling uploading or scheduling; it’s a matter of how much you want to set up. People generally dislike AI videos, but well-done ones are absolutely doable. This workflow with or without AI is also useful for learning automation, which is not the same as AI, and will open you to a wider world of tools.
Tip: If your content uses AI-generated voice, images, or music, TikTok requires you to toggle the AI-generated content disclosure when uploading. This is good practice and may be required by platform policy.
Toolkit Reference
Below are the plugins, extensions, and MCP servers used across the steps in this guide.
Community Plugins
- Everything Claude Code
- Agent workflow optimization, skill system, and development patterns
- Codex Integration
- Code review via OpenAI Codex CLI
Official Plugins
These are built-in to Claude Code CLI and can be installed directly from the plugin menu.
- superpowers
- Planning, brainstorming, and implementation workflows
- context7
- Documentation lookup for Resolve API, ffmpeg, and ComfyUI
- firecrawl
- Web research for finding tools and trending topics
- serena
- Semantic code analysis for automation scripts
- code-review
- Automated code review for Python scripts
- code-simplifier
- Code cleanup and refactoring
MCP Servers
- brave-search
- Web search for research and finding tools
Integrations
- GitHub
- Source code, open-source projects, and model repositories