comfyui-h3-editor
A timeline video editor for MiniMax H3 inside ComfyUI: write shots, render them on your own GPU, cut picture and sound together, or let an agent make the video.
Safety scan
A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.
Check it deeper on your machine
It has changed since its scan on 11 Nov 2026.
As a Claude Code plugin, install it and then ask Claude Code:
/plugin marketplace add keithmackay/mackayi/plugin install scanskillsafety@mackayiDo a deep scan of https://github.com/igorls/comfyui-h3-editor with scanskillsafety
Or from a terminal, clone it and run it:
git clone https://github.com/keithmackay/scanskillsafety-skill ~/.claude/skills/scanskillsafetynode ~/.claude/skills/scanskillsafety/cli.cjs --deep 'https://github.com/igorls/comfyui-h3-editor'
It checks the exact commit: every file, its packages against OSV (api.osv.dev), how it compares with this page's scan (findsafeskills.com receives the repository's address), and, if you agree, Claude reading every file with your own Claude subscription. It never runs the code.
Badge for your README
Markdown
[](/listings/igorls--comfyui-h3-editor)HTML
<a href="/listings/igorls--comfyui-h3-editor"><img src="/api/listings/igorls--comfyui-h3-editor/badge.svg" alt="findsafeskills safety rating" /></a>Install
Install instructions from the README
As written by the repo's author. findsafeskills has not verified these steps.
1. Install
You need a recent ComfyUI with its built-in MiniMax H3 support. In ComfyUI's `custom_nodes` folder:
git clone https://github.com/igorls/comfyui-h3-editor
pip install -r comfyui-h3-editor/requirements.txt
On the Windows portable build, run the second step as `..\python_embeded\python.exe -m pip install -r comfyui-h3-editor\requirements.txt`. The requirements bring ffmpeg, for importing and exporting video, and the gguf reader for HyperFlow's models.
Restart ComfyUI. **H3 Editor** appears in the sidebar.
2. Download the models
The pack does not download models. These are the ones a first video needs, about 52 GB:
| File | Folder | Size | Where to get it | |---|---|---|---| | `qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors` | `models/text_encoders` | 15.7 GB | [Comfy-Org/MiniMax-H3](https://huggingface.co/Comfy-Org/MiniMax-H3), `text_encoders/` | | `minimax_h3_video_vae_fp16.safetensors` | `models/vae` | 5.2 GB | Comfy-Org/MiniMax-H3, `vae/` | | `minimax_h3_audio_vae_fp32.safetensors` | `models/vae` | 0.6 GB | Comfy-Org/MiniMax-H3, `vae/` | | `minimax_h3_ref2va_hyperflow_flash_14b-Q8_0.gguf` | `models/diffusion_models/hyperflow` | 15.2 GB | [VideoRebirthAI/hyperflow-h3-flash](https://huggingface.co/VideoRebirthAI/hyperflow-h3-flash), `ref2va_gguf/` | | `minimax_h3_fl2va_hyperflow_flash_14b-Q8_0.gguf` | `models/diffusion_models/hyperflow` | 14.9 GB | VideoRebirthAI/hyperflow-h3-flash, `fl2va_gguf/` |
HyperFlow's smaller quantizations, from Q3_K_M at 7.2 GB, work too, anywhere in `models/diffusion_models`.
These are optional, and worth having:
| File | Folder | Size | Where to get it | For | |---|---|---|---|---| | `minimax_h3_latent_upscaler_3d_conv_v1_bf16.safetensors` | `models/latent_upscale_models` | 0.7 GB | [LBH-123-AI/Minimax_h3_latent_Upscaler](https://huggingface.co/LBH-123-AI/Minimax_h3_latent_Upscaler) | Finals at up to 1920 pixels | | `MMH3-RealismSlider-V1.safetensors` | `models/loras` | 3 MB | Civitai model 3014078 | Natural skin in finals | | `taeh3.safetensors` | `models/vae_approx` | 23 MB | [OzzyGT/taeh3](https://huggingface.co/OzzyGT/taeh3) | The moving live preview |
Other H3 checkpoints work as well; see [models and quality](docs/models.md).
3. Make your first video
1. Open **H3 Editor** from the ComfyUI sidebar. It starts you on a first project. 2. In the **Scene** panel, choose the format: 16:9, 9:16 or 1:1. 3. Press **New clip**, write what happens in **What happens**, and press **Render draft**. The viewer shows the clip taking shape. 4. When the draft lands, press Space to play it. Not right? Change the text and render again: each render becomes another take. 5. Press **Extend forward** to carry the shot on, or add another clip. 6. Press **Render final** for the takes you keep, then **Export**. The video is saved in ComfyUI's output folder, under `h3take_projects/<project>/exports`.
What was scanned
The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.
README.md · README · 436 lines
View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)
This file has changed since it was scanned, so this is not exactly the text that was scanned, and no line is marked. The findings are listed with the safety scan above.
- # H3 Editor
- A timeline video editor for MiniMax H3, inside ComfyUI. Write your shots,
- render them on your own GPU, and cut them together, picture and sound.
- **[Project site](https://igorls.github.io/comfyui-h3-editor/)**: watch a cut
- made with it, shot by shot.
- 
- ## What you can do
- - **Write shots, not graphs.** Describe what happens and what is said. The
- editor builds the ComfyUI graph, queues it, and shows the clip moving, with
- sound, while it samples.
- - **Keep people and places the same.** Add their pictures to the cast, up to 8
- per member, and write @Name in a shot or drag the member into the text. Give
- them a voice, or add a narrator who is only a voice.
- - **Join shots without a visible cut.** Extend a clip forward or back, bridge
- two clips, or merge a join. A new clip carries on from real frames and sound
- of the clips it joins.
- - **Edit as in any editor.** Trim, split, move, change speed, add transitions,
- flip between takes, file clips in bins, bring in your own footage and
- pictures, and export one video.
- - **Add music and narration.** Lay sound files under the cut, trim them,
- fade their ends and draw volume curves under dialogue. The export mixes them in.
- - **Set how loud each clip is.** A clip on the timeline has its own volume,
- from silent to twice as loud, and the export applies it. The viewer has a
- volume of its own beside the play controls, which changes only what you hear.
- - **Make a game sprite (experimental).** In an independent shot, open **Sprite Draft**, choose a facing, camera, canvas (**Square 512×512**, **Portrait 256×384** or **Landscape 384×256**) and clip length (**39**, **56** or **73 frames**), and select a jog, a planted idle, the shot prompt or an imported motion guide. Tag cast pictures by view (**front**, **side**, **back**, **left**) so directional renders route only heading-matched plates. Sprite Draft prefers installed HyperFlow H3 Flash, then H3 Turbo, skipping audio decode on drafts. Auto cutout uses BiRefNet with `H3DefringeMatte` edge defringing when installed in ComfyUI’s `background_removal` model folder and otherwise uses the background key. Each take caches lossless RGBA frames so **Export size** (**64**, **128**, **160**, **256 px** or any even width × height in `32..512`) and **24 fps** re-export without re-decoding video. **Close loop** pins a clean mid-stride frame at `0`, `stride` and `2×stride`. **Finish draft** refines the selected motion with the same loop and pivot. **Props & plates…** renders 4+ MP static props, paired state variants (`unlit → lit`) and character turnaround plates with P-H3-Image.
- - **Try fast, finish large.** Drafts are small and quick. Finals are refined at
- up to 1920 pixels, with a realism LoRA for natural skin.
- - **Pick the model, per project or per clip.** HyperFlow H3 Flash, MiniMax H3,
- its Turbo and pi-H3 LoRAs, Singularity, DaSiWa, 10Eros Max and PRISM are built in, and a clip
- can render on a different one than the rest of the project. Any other H3
- checkpoint works too: pick its file, set its steps, sampler and scheduler,
- and save it as a preset for every project.
- - **Plug in speed-ups from other packs.** Sparse attention, a step cache or a
- compile node: any installed node that takes the model and returns it can go
- on every render of a project, with its own settings shown in the editor.
- - **Let a language model write.** Describe the video and it writes the shots as
- clips, or rewrites one clip from what you ask for. It is sent the cast's
- pictures, and any picture you add, such as a place or a frame of the cut, so
- a model that can see writes from them. It uses the model you name: a text
- model run by ComfyUI itself, or any OpenAI-compatible endpoint, Anthropic,
- Gemini or Ollama.
- - **Bring your LoRAs.** Add LoRAs trained on MiniMax H3 to a project, or to one
- clip, set each one's strength, and choose whether it runs on every step or
- only after the first ones have set the framing.
- - **Generate to a recorded voice.** Give a clip a recorded track, such as an
- approved narration, and its picture is generated to it.
- - **Let an agent make the video.** Everything the editor does is an HTTP call,
- and an agent skill teaches the whole workflow.
- - **RefMods, if you use them.** With the ComfyUI-MiniMaxH3Mod pack installed, a
- cast member can carry a RefMod, and the editor can make one from the member's
- pictures.
- - **Experimental VOSR2 upscaling.** Install the optional ComfyUI-VOSR2 pack,
- then select VOSR2 under **Upscale 2x** on a final take or imported footage.
- The editor keeps the original sound and saves the result as another take.
- See [upscalers](docs/models.md#upscalers) for setup and model downloads.
- Everything else needs no other custom node pack.
- ## Quick start
- ### 1. Install
- You need a recent ComfyUI with its built-in MiniMax H3 support: 0.43 or newer.
- The pack is tested on 0.44.
- **From the Comfy Registry.** Search for **H3 Editor** in ComfyUI-Manager, or
- use comfy-cli:
- ```bash
- comfy node install h3-editor
- ```
- This installs the pack into `custom_nodes/h3-editor`, with its requirements.
- **From GitHub.** In ComfyUI's `custom_nodes` folder:
- ```bash
- git clone https://github.com/igorls/comfyui-h3-editor
- ```
- ```bash
- pip install -r comfyui-h3-editor/requirements.txt
- ```
- On the Windows portable build, run the second step as
- `..\python_embeded\python.exe -m pip install -r comfyui-h3-editor\requirements.txt`.
- The requirements bring ffmpeg, for importing and exporting video, and the gguf
- reader for HyperFlow's GGUF models.
- Restart ComfyUI. **H3 Editor** appears in the sidebar.
- Keep one copy only. The two routes use different folders, `h3-editor` and
- `comfyui-h3-editor`, and ComfyUI would load both. To move from a GitHub clone to
- the registry, delete `custom_nodes/comfyui-h3-editor` before you install. Your
- projects and settings are not in that folder: projects live in ComfyUI's
- `output` folder and settings in its `user` folder, so they stay.
- ### 2. Download the models
- Open **H3 Editor** from the ComfyUI sidebar. When the models are missing, it
- opens **Get the models**, which lists the files below with their sizes and
- licenses: press **Download** and each one lands in its folder. The latent
- upscaler for finals comes too. It never replaces a file you already have, and
- it carries on where it stopped if you stop it or the connection breaks. See
- [get the models from the editor](docs/models.md#get-the-models-from-the-editor).
- To put them in place yourself instead, these are the ones a first video needs,
- about 54 GB:
- | File | Folder | Size | Where to get it |
- |---|---|---|---|
- | `qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors` | `models/text_encoders` | 15.7 GB | [Comfy-Org/MiniMax-H3](https://huggingface.co/Comfy-Org/MiniMax-H3), `text_encoders/` |
- | `minimax_h3_video_vae_fp16.safetensors` | `models/vae` | 5.2 GB | Comfy-Org/MiniMax-H3, `vae/` |
- | `minimax_h3_audio_vae_fp32.safetensors` | `models/vae` | 0.6 GB | Comfy-Org/MiniMax-H3, `vae/` |
- | `minimax_h3_ref2va_hyperflow_flash_14b-int8_convrot.safetensors` | `models/diffusion_models/hyperflow` | 16.4 GB | [VideoRebirthAI/hyperflow-h3-flash](https://huggingface.co/VideoRebirthAI/hyperflow-h3-flash), `ref2va_int8_convrot_comfyui/` |
- | `minimax_h3_fl2va_hyperflow_flash_14b-int8_convrot.safetensors` | `models/diffusion_models/hyperflow` | 16.0 GB | VideoRebirthAI/hyperflow-h3-flash, `fl2va_int8_convrot_comfyui/` |
- HyperFlow's model page recommends these INT8 files on RTX GPUs with 24 GB or
- more. On a smaller GPU, take its GGUF files in their place, from Q8_0 at 15.2 GB
- down to Q3_K_M at 7.2 GB, anywhere in `models/diffusion_models`; see
- [models and quality](docs/models.md#hyperflows-other-files).
- The text encoder and the video VAE also come in other precisions in the same
- repository: `qwen3vl_32b_minimax_h3_int8_convrot.safetensors` (25 GB) and
- `qwen3vl_32b_minimax_h3_bf16.safetensors` (48 GB), and
- `minimax_h3_video_vae_int8_convrot.safetensors` (2.6 GB). Any one of each is
- enough: the editor renders with what it finds, and with more than one installed
- you pick in the Project tab. A Qwen3-VL 4B or 8B can also stand in for the 32B text
- encoder, in about a third of the memory: see
- [a smaller text encoder](docs/models.md#a-smaller-text-encoder).
- These are optional, and worth having:
- | File | Folder | Size | Where to get it | For |
- |---|---|---|---|---|
- | `minimax_h3_latent_upscaler_3d_conv_v1_bf16.safetensors` | `models/latent_upscale_models` | 0.7 GB | [LBH-123-AI/Minimax_h3_latent_Upscaler](https://huggingface.co/LBH-123-AI/Minimax_h3_latent_Upscaler) | Finals at up to 1920 pixels |
- | `MMH3-Realism-V2.safetensors` | `models/loras` | 71 MB | [Civitai model 3041489](https://civitai.com/models/3041489) | Natural skin in finals. Civitai asks you to sign in, so the editor cannot download it for you |
- | `taeh3.safetensors` | `models/vae_approx` | 23 MB | [OzzyGT/taeh3](https://huggingface.co/OzzyGT/taeh3) | The moving live preview |
- Other H3 checkpoints work as well; see [models and quality](docs/models.md).
- ### 3. Make your first video
- 1. Open **H3 Editor** from the ComfyUI sidebar. It starts you on a first project.
- 2. In the **Project** tab on the right, choose an aspect ratio or enter your own.
- **Draft size** and **Final size** each offer **Your own size**, where you can set
- width and height directly or calculate them from a longest side or megapixel target.
- 3. Press **New clip**, write what happens in **What happens**, and press
- **Render draft**. The viewer shows the clip taking shape.
- 4. When the draft lands, press Space to play it. Not right? Change the text and
- render again: each render becomes another take.
- 5. Press **Generate**, then **Extend forward**, to carry the shot on, or add another clip.
- 6. Press **Render final** for the takes you keep, then **Export**. The video is
- saved in ComfyUI's output folder, under
- `h3take_projects/<project>/exports`.
- ## Using the editor
- **The panel on the right.** Select a clip and it shows the clip in tabs:
- **Shot** for what happens, the cast, sound and model, **Takes** for its renders,
- and **Timeline** for its speed and transition. The **Project** tab is always
- there, with the look, format, model and LoRAs that every clip shares. Esc
- selects nothing.
- **Clips and takes.** Every render of a clip is a take with its own seed and the
- prompt that made it. Run the pointer across a take's picture to look through
- it, flip between takes with `[` and `]`, keep a seed to render variations of
- one result, or restore a take's prompt, style markers and mid-frame guides. To cut between takes,
- expand the clip in **Clips** and drag individual takes onto the picture track.
- Click a take in that list, or choose **Add as clip** in the inspector, to append
- it to the cut. Each placement creates an independent clip using the existing
- render, so choosing a different take on the original clip leaves the cut intact.
- In **Takes → Manage takes**, select the versions to remove. You can clear every
- render from a generated shot and render it again with the same prompt and edits.
- Removing the active take selects the newest one left. Media files, independently
- placed takes and earlier revisions stay available.
- Rendered clips stay on disk: moving,
- trimming or splitting them never renders anything again. **Render all drafts**
- renders every clip that has no take yet, the ones that share a model one after
- another.
- **Mid-frame guides (experimental).** Select a generated shot and open **Shot → Mid-frame guides**.
- Move the playhead inside the shot and choose **Upload image** or **Use playhead frame**.
- You can also enter the destination source frame explicitly, including before the first render.
- The list shows each picture and its zero-based frame. Change the frame or remove the picture
- before generating another draft or final. All placements of the shot share its guides.
- Check the motion approaching and leaving each guide. Some checkpoints can pause or jump near it.
- **Cast and voices.** **Add** in the Cast panel adds a reference picture or a
- voice. Choose who is in each clip, and write @Name in **What happens** to place
- them, or drag their card into the text. Write @Name in **Who speaks** to use
- their voice. A voice can be uploaded, or taken from a clip on the timeline.
- **RefMods.** A RefMod is a reference saved as one file by the
- ComfyUI-MiniMaxH3Mod pack. With that pack installed and the file in ComfyUI's
- `models/refmods` folder, there are two ways to use one you already have. **Add**,
- **Add RefMod** in the Cast panel makes a cast member from a RefMod alone, with
- no pictures to upload. Or open a cast member and choose the file under
- **RefMod**, with its strength. The menu lists what the pack's **Load H3
- RefMods** node lists, subfolders included; reload the editor after adding a file,
- and restart ComfyUI if it still does not appear. A member that is only a RefMod renders on the prompt-only
- model, so describe them in the member's description. A member with pictures can
- also get a RefMod made from them: **Make a RefMod from these pictures** in the
- same section.
- **The cast catalog.** Press **Save** beside **Catalog** on a cast member to keep
- them for every project, with their pictures and voice. In another project,
- **Add**, **From the catalog** brings in a copy. When you change someone and save
- them to the catalog again, the other projects show **Newer in the catalog**
- with an **Update** button, or you can bring every project up to date at once from
- the catalog. A copy you changed in its project is never overwritten without you.
- **Joining shots.** Select a clip on the timeline, open **Generate** and choose **Extend forward** or
- **Extend back** to generate what happens next or before. **Bridge to next**
- generates what leads from it into the following clip, landing on a short run of
- its first frames, and **Bridge into its movement** holds more of them, for a
- following clip that starts in mid-movement. **Merge with next**
- replaces the join between two clips with a generated crossing. These clips hold
- 39 real frames and their sound from the clips they join, so the cut does not
- show. If a clip they join changes, they are marked: render them again. An empty
- space between two rendered clips shows a plus that generates the bridge.
- **First and last frames.** Give a clip an exact first or last frame by
- dropping a picture on it, uploading one, or using the frame at the playhead,
- with the buttons beside each frame in the Shot tab.
- **Picture to a track.** A clip can hold a recorded sound, and its picture is
- generated to it. Extend forward carries on with the next part of the same
- track, so a long talking performance can be built from short shots.
- **Your own footage.** **Import** brings in a video or a picture, fitted to the
- project's format. You can trim it, and extend or bridge from it like a render.
- **Edit with prompt.** Choose this in **Generate**, the clip's context menu, or
- **Takes** on imported footage or a generated clip. It uses the marked In/Out
- range or the clip's current trim (2–15 seconds). Describe the change and
- optionally supply a replacement first frame. The result appears under **Takes**
- as a full-length alternate with the original sound and the picture outside the
- edit range intact.
- **Motion transfer.** Choose **Motion transfer** in **Generate**, the clip's
- context menu, or **Takes** on imported footage or a generated clip, and select
- 2–30 seconds of one performer. Pick a cast member or upload a character image,
- then choose **Replace character in scene** (masked replacement on the original
- background using SAM 3.1, Character Swap LoRA, and zero-halo plate compositing,
- with **Keep reference style** for stylized/anime characters or **Match scene
- lighting** for photorealistic people) or **New scene from pose** (pose-guided
- choreography via RT-DETR, gated SDPose, and H3 Fun ControlNet Union 2.0). Clips
- longer than 5.9 seconds automatically tile into seamless 124-frame latent-locked
- relay windows (`22`-frame overlap) in a single graph so VRAM stays constant.
- The result is a normal clip placed right after the source on the timeline with
- the exact source audio, so you can trim, split, extend, add VFX, or re-render
- Draft → Final / Sharp 2× from the inspector. Missing optional models download
- from Hugging Face with sha256 verification right inside the Motion transfer
- dialog; see [motion transfer models](docs/models.md#motion-transfer-models).
- **Compare takes.** In **Takes**, click **Compare takes**, Ctrl-click or
- Command-click a second take, or drag it into the viewer. Play or step
- through both versions together and drag the divider to reveal either
- side. Audio plays from take A. **Done** closes the comparison.
- **Music and narration.** **Add sound** on the timeline lays a sound file under
- the cut at the playhead, on a lane of its own. Drag it to move it, up or down to
- change its lane, and drag its ends to trim it; its volume is in the panel on
- the right. In **Volume curve**, set the fade lengths in frames. Move the
- playhead and press **Add point at playhead**, then drag the point on the waveform
- or type its level in the panel. For example, add four points at 100%, 25%, 25%
- and 100% to lower the music under a line of dialogue and bring it back up.
- Points follow their source audio when you trim or move the sound. The export
- mixes it under the clips' own sound.
- **Notes on frames.** Stop on a frame and press **N**, or **Note** under the
- viewer. Pin a spot, drag a box or draw on the picture, and write what should
- change. With a range marked by I and O, the note covers the range. Notes show
- on the ruler and in the Project tab, stay on their frame when clips move, and
- are how you tell an agent exactly what to fix: it reads each note with a
- picture of its frame.
- **Checks.** The editor measures every rendered cut from its pictures: jump
- cuts, takes that cut, fade, freeze or go still inside, clips that are soft
- next to the rest, and extensions that drift from where their shot began. Each
- finding is marked on the timeline, and the chip at the top counts them and
- lists them, worst first; click one to go to it.
- **Right-click.** A clip on the timeline or in the Clips panel, a VFX clip, a
- sound, a cast member and the viewer each have a menu of what can be done to
- them, such as render, rewrite, extend, split or use the frame as a first frame.
- **Help in the panel.** The **?** at the end of the panel's tabs shows the
- explanations under each setting; with it off, each one is the setting's tip.
- **The Clips panel.** Click a clip to play it on its own, whether or not it is on
- the timeline; click the timeline to go back to the cut. File clips in bins with
- the folder button and drag clips onto them. Delete a clip with the trash button
- in its panel, or with Delete in the Clips panel.
- **Playback.** The **Loop** button beside Play switches between playing once and
- repeating. **In** and **Out** mark the range to repeat; Out includes the frame on
- screen. Clear the marks to repeat the whole cut. Clip previews have their own
- range and can loop too. The **Sound** menu above the timeline switches all sound
- lanes between Waveform and Spectrogram.
- **Crop and reframe.** Select a placement on the picture track, open its **Timeline**
- tab and choose **Crop and reframe**. Drag the box around the subject and use its
- corner or **Zoom** to resize it. **Apply crop** affects this placement in the
- viewer and export. **Reset framing** restores the full source at the cut's aspect ratio.
- **Export a range.** Mark **In** and **Out** on the timeline and click **Export**.
- The dialog selects **Marked range** and shows its duration. You can also choose
- **Whole timeline**. The exported range includes the displayed Out frame, picture
- transitions, rendered VFX and the mixed soundtrack.
- **Reusable looks.** The **Style library** in the Project tab includes eight
- built-in looks. **Save look** keeps your own style for every project on this
- ComfyUI, in `user/h3take/looks.json`, outside project folders. Applying a look
- copies its text into the project's style, where you can edit it.
- **Styles within a shot (experimental).** Put the playhead inside a generated
- clip and press **Style marker**. Choose looks and frame positions in the clip's
- Timeline tab. Each marker is the destination of a smooth blend from the
- previous look. Render the clip again to see it; these are generation controls.
- See [style transitions](docs/style-timeline.md) for the node setup, cost and limits.
- **Batch renders.** The header's **Render** menu renders all missing drafts or
- all clips whose selected take is not a final. It also cancels all renders, or
- only finals, in the current project. **Finish draft** can redraw a small draft
- at Large, Medium or a custom final size as well as Full; it needs the latent
- upscaler. Sharp uses a separate video-upscaling path and needs a new final.
- **Models and LoRAs.** **Models and their settings** in the Project tab lists
- every model the project can render with, and adds your own checkpoints. The
- LoRAs section adds LoRAs for the whole project; a clip's Shot tab has **LoRAs
- for this clip**. **Model patches** puts nodes from your other packs on the
- model of every render. See [models and quality](docs/models.md).
- **Projects.** Projects live in `h3take_projects` in ComfyUI's output folder.
- The menu beside the project list renames, duplicates or deletes the open one; a
- deleted project is moved to a `.trash` folder there, never erased. **Download
- as zip** in the same menu packs a project into one file, with every take, only
- the takes the cut uses, or no renders, and **Open a project zip** opens it as a
- new project, here or on another machine such as a cloud GPU. Renders run
- on the server, so closing the editor never loses one. Every change is a
- revision you can go back to, and two editors, or an editor and an agent, can
- work on one project at once.
- **Project storage.** Open the menu beside the project name and choose **Project
- storage** to see how much space its `.hold.safetensors` files use. **Clear unused
- files** removes cached latents the current project no longer references. Choose
- **All cached latents** to clear those for existing takes too. Your videos still
- play and export. Extensions can use frames from the video, and you can render a
- new draft when you need its original latents for finishing.
- | Key | Does |
- |---|---|
- | Space | Play or pause |
- | I, O | Mark In and Out; enable Loop to repeat the range |
- | Left, Right | Move one frame. Hold Shift for one second |
- | Up, Down | Jump to the previous or next edit |
- | Home, End | Go to the start or end |
- | S | Split at the playhead |
- | Esc | Select nothing |
- | Tab, Enter | Reach a clip or a sound on the timeline, and select it |
- | Delete | On the timeline, remove a clip from it, or close a selected gap. In the Clips panel, delete the clip or bin |
- | Shift+Delete | Remove a clip from the timeline and close the space |
- | [ , ] | Flip between takes of the selected clip |
- | , . | Slip the selected clip: show an earlier or later part of its take without moving it. With Shift, half a second |
- | Alt+drag | Slip a clip on the timeline |
- | Ctrl+wheel, \\ | Zoom at the pointer, fit the whole scene |
- | Ctrl+Z, Ctrl+Shift+Z | Undo, redo |
- Press ? in the editor for this list.
- ## Going further
- - **[Models and quality](docs/models.md):** every checkpoint the editor can
- render with, how finals are refined, how LoRAs land, and model patches.
- - **[The nodes](docs/nodes.md):** H3 Take and the other nodes, for building
- your own ComfyUI workflows.
- - **[For agents](skills/h3-editor/SKILL.md):** a skill that teaches an agent to
- plan shots, render, review and export through the editor's HTTP API, with
- `scripts/h3.py`, a client that needs only Python. Copy `skills/h3-editor` into
- an agent's skills folder to install it. The full API is in
- [references/api.md](skills/h3-editor/references/api.md).
- ## Troubleshooting
- | You see | Do |
- |---|---|
- | A model in the Model menu says "not installed" | Its hint names the missing file, and so does the Models dialog. Put it in the folder from the tables above, then reload the editor |
- | "ffmpeg is missing" | Install the requirements, step 1 above, and restart ComfyUI |
- | "HyperFlow H3 Flash GGUF files need the gguf Python package" | The same |
- | Finals come out at 768x1344 or 1344x768 | Add the latent upscaler. The Model hint says so too |
- | The live preview is a still | Add `taeh3.safetensors` |
- | ComfyUI runs out of memory | Pick a smaller **Final size** in the Project tab, try one of HyperFlow's GGUF files, which come in smaller sizes, use [a smaller text encoder](docs/models.md#a-smaller-text-encoder), or render the finals on a Colab GPU with `colab/h3_editor.ipynb`. So far the editor has been tested on one 96 GB GPU only |
- ## Sprites workspace
- Choose **Sprites** beside **Video** to review a character's animations. Add a reference sheet or use an existing cast member, and tag each cast photo's view (**front**, **side**, **back**, **left**) in **Edit references** so directional drafts automatically route heading-matched plates. Choose the camera, canvas (**Square 512×512**, **Portrait 256×384** or **Landscape 384×256**), clip length (**39**, **56** or **73 frames**), four or eight directions, and a motion. Click a direction to compare its takes in sync. Focus the game preview and move with WASD, the arrow keys, or by holding a point on the ground.
- Choose **Add animation…** to name a loop and describe the movement once for this character. Select a direction and click **New draft**. Each direction uses the shared motion prompt, heading-matched character references and selected camera. **Edit prompt…** changes the description for subsequent drafts. Existing takes retain their original prompt. The atlas lists each named animation by direction, such as `Wing stretch_NE`.
- **New draft** queues one seed for the selected direction. Click a take to use it for that slot. **Preview mirrored** reflects the opposite direction. Click **Use mirrored** to keep it. **Close loop** extracts a clean non-edge-touching mid-stride frame from the selected take and queues a self-pinned draft (anchoring frames `0`, `stride` and `2×stride`) for seamless walk/run closure. **Edit take** opens the source shot in Video, and **Open in Sprites** returns from its Sprite Draft controls. **Finish this take** refines the selected draft.
- Choose **Props & plates…** to queue ≥ 4 MP **P-H3-Image** renders (`2048×2048`, `1664×2496` or `2496×1664`) with BiRefNet cutout and `H3DefringeMatte` edge defringing: style-locked static game props (with optional 42° top-canopy framing), pixel-aligned paired state variants (`unlit → lit`), or neutral standing character turnaround plates. It needs a ComfyUI with the `PH3ImageConditioning` node and these files: `p-h3-image_int8_convrot.safetensors` or `p-h3-image_bf16.safetensors` in `diffusion_models` (or `p-h3-image_lora_bf16.safetensors` in `loras`, applied to `minimax_h3_ref2va_pruned_int8_convrot.safetensors`), `p-h3-image_text_encoder_int8_convrot.safetensors` or the H3 text encoder in `text_encoders`, and `p-h3-image_vae_bf16.safetensors` in `vae`.
- **Export atlas** packs available loops for that character and camera into a ZIP containing `sprites.png`, optional `sprites.webp` (**Include WebP**, on by default), an Animwright-format `sprites.json`, and an offline `preview.html`. Each loop records its frame rectangles, ground anchor, frame rate, source take, seed and mirror choice. Export size (**64**, **128**, **160**, **256 px** or any even **Custom…** size / `W×H` in `32..512`), output style and frame rate apply to the whole atlas, re-packing from each take's lossless RGBA frame cache when present. **Original per loop** keeps each motion’s timing. Choose **24 fps** to include intermediate frames from the saved takes at the same playback speed. A seven-pose run becomes 14 frames, and a two-second idle becomes 48 frames.
- ### Test an atlas without installing H3 Editor
- Open [the sprite test room](editor/sprite-test.html) in a browser from a checkout or downloaded copy of this repository. Its built-in calibration marker moves in eight directions. Choose **Open PNG and JSON** to test your own exported atlas, or extract an H3 sprite ZIP and open its included `preview.html`. The room runs offline without a game, GPU, model weights or server. Switch between checkerboard, light and dark grounds to inspect the cutout, and turn on the pivot to check ground contact at native size.
- ## Develop
- `python -m unittest discover -s tests` runs from the repo root. The writer,
- LoRA and HyperFlow tests need torch and ComfyUI, with `PYTHONPATH` set to
- ComfyUI's folder, and the GGUF tests need gguf too. They skip without them.
- `node --test tests/editor.test.mjs` checks playback clocks, batch actions and
- spectrogram trimming without browser dependencies.
- `tools/deploy.sh` copies the repo into a running container, runs the tests
- there, and reloads the pack in place, so models stay loaded. Refresh the browser
- afterwards. `--restart` restarts ComfyUI instead. `tools/run_take.py` queues a
- real take, and `tools/compare_takes.py` measures how closely a span matches its
- source.
- ## Credits and license
- GNU General Public License v3.0, in [LICENSE](LICENSE). Parts are adapted from
- ComfyUI-HyperFlow-Flash, ComfyUI-GGUF and ComfyUI-MiniMax-H3-Turbo, and keep
- their own licenses; [NOTICE](NOTICE) lists them. Model files come under their
- own licenses, which this pack's license does not change.
Ratings and comments
Sign in to rate this skill.
Sign inFree to use. If it’s helped, pass it on.