gif-maker

SkillMedia

Short looping GIF utility. Modes: (A) MP4-to-GIF with 2-pass palette optimization; (B) generate 1-3s clip via Veo/Sora/Kling/Runway/fal-video then convert. Aspect presets 1:1, 9:16, 16:9, 2:1. Use when: 'gif', 'looping animation', 'short loop', 'make a gif', 'сделай гифку', 'короткая зацикленная анимация'.

Available today. Use it from your connected AI after setup.

Add ahel to your AI once: Claude, ChatGPT, Cursor, Claude Code or Codex. Then ask it to use this.

Then ask your AI: use the gif-maker skill

What this skill tells your AI

The instructions your AI receives, as published by mikefluff/skills in skills/gif-maker/SKILL.md and read by ahel’s review.

Distinct from reel-builder:

  • No script / music / captions — single short looping clip
  • Output format is GIF, not MP4
  • Much shorter duration (1-3s vs 8-30s)

Distinct from video-prompt:

  • Output is GIF (palette-optimized), not MP4
  • Wraps video-prompt --execute for generation mode, then converts
  • Aspect crop and frame-rate optimization built-in

This skill does NOT:

  • Generate long videos (>5s) — those are reels / shorts; use reel-builder
  • Output WebP / APNG / animated AVIF — GIF only (most universally supported format)
  • Background-removal frame-by-frame — GIF supports binary transparency but skill doesn't try
  • Combine multiple clips into one GIF — single source → single GIF

ROLE

Two modes:

Mode A — convert MP4 to GIF: Read an existing MP4 → apply optional aspect crop + trim → run 2-pass palette generation (palettegen + paletteuse) for high-quality color → save GIF.

Mode B — generate then convert: Read prompt + provider → call video provider for a short clip → save intermediate MP4 → run conversion pipeline → save GIF.

PIPELINE

Mode A — gif-maker --input <mp4>

  1. Resolve input — path to existing MP4.
  2. Detect ffmpeg — fail if absent (print install instructions).
  3. Aspect crop (optional) — center-crop to --aspect 1:1|9:16|16:9|4:5|2:1|1:2.
  4. Trim (optional) — --start <s> --duration <s>.
  5. 2-pass palette generation:
    • Pass 1: palettegen → palette.png (256 colors, stats_mode=diff)
    • Pass 2: paletteuse with bayer dithering → output GIF
  6. Save to ./generated/gif/<stem>.gif (or --output <path>).

Mode B — gif-maker --prompt "..." --model <video-slug>

  1. Resolve prompt + model (default suggestion: veo-3-1-fast for cheapest, kling-3 for best motion).
  2. Pre-flight: check provider available + cost-confirm.
  3. Generate — call video provider (poll if async).
  4. Save intermediate MP4 to ./generated/gif/_source/.
  5. Run Mode A pipeline on the intermediate.
  6. Final GIF to ./generated/gif/<slug>.gif.

MODES

Mode A — convert existing

  • gif-maker --input <path.mp4> (required)

Mode B — generate then convert

  • gif-maker --prompt "<text>" --model <video-slug> (both required)
  • --prompt-file <path> — alternative to --prompt
  • --duration <seconds> — target generation length (default depends on model, typically 3-5s; we trim down to keep GIF small)

Output controls (both modes)

  • --output <path> — explicit output GIF path
  • --aspect 1:1|9:16|16:9|4:5|2:1|1:2 — center-crop to ratio
  • --fps N — GIF frame rate (default 12; range 6-24)
  • --width N — output width in px (default by aspect: 720 for landscape, 540 for portrait/square)
  • --start <s> — trim start
  • --duration <s> — trim length (Mode A) / target generation (Mode B)
  • --yes — skip cost confirmation (Mode B)
  • --timeout <s> — poll timeout for async video gen (default 600)

REFERENCES (load on demand)

FileWhen to load
references/quality-tuning.mdWhen GIF is too large, colors look banded, or motion is choppy
references/model-picker.mdMode B — which video provider for which kind of clip (motion / talking head / abstract loop)
references/troubleshoot.mdWhen ffmpeg fails, palette breaks, source is too long, etc.

EXAMPLES

See examples/before-after.md — 3 calibration runs: convert existing reel to social GIF, generate abstract loop from prompt, convert long video to short looping GIF with trim.

CONSTRAINTS

  • GIFs should be SHORT. 1-3 seconds is the sweet spot. 5+ seconds creates huge files (>10MB), losing the "lightweight" advantage of GIF over MP4.

  • GIFs are large by nature. Even with palette optimization, expect 500KB-3MB for 2-3 second clips. For smaller files: lower --fps (8-10), lower --width (480 or smaller), shorter --duration.

  • Lossy color via 256-color palette. GIF max is 256 colors. Photoreal video → noticeable banding. Animated motion / illustrations / cartoons → great. Cue this in Mode B prompt: "flat illustration / animated / cartoon aesthetic" works better than "photoreal".

  • Aspect crop is center-crop. Source aspect 16:9, target 1:1 → loses left+right portions.

  • Looping is implicit. GIFs always loop by default (loop=0 = infinite). Source video should naturally loop or close-to-loop for best effect.

  • No audio. GIFs don't support audio. If user needs audio + animation, that's MP4 territory (use reel-builder or video-prompt --execute).

  • ffmpeg required. Install via brew install ffmpeg (Mac) / apt-get install ffmpeg (Debian). Without it, skill prints the manual ffmpeg command for the user to run.

  • Default fps 12. 12fps is the GIF sweet spot. Higher fps = bigger file. Lower fps = choppy motion. Override only when needed.

  • Cost in Mode B depends on the chosen video provider. Veo 3.1 Fast ~$0.40/sec; Kling 3 ~$0.50/sec; Sora 2 ~$1.00/sec. A 3-second clip = $1.20-$3.00.

  • No frame interpolation. This skill doesn't upsample frame rates — it samples the existing video at --fps. For smoother motion, generate at higher native fps in Mode B.

  • Never print API keys (Mode B). Mask in errors.

INVOCATION HINTS

When the user says any of:

  • "gif", "make a gif", "looping animation", "short loop"
  • "convert this MP4 to GIF", "compress this video as a GIF"
  • "сделай гифку", "короткая зацикленная анимация", "конвертируй в gif"
  • "loop of X", "2-second loop of Y for Twitter"

Mode A is the default when --input is provided. Mode B requires both --prompt and --model.

Defaults: --fps 12 --width <by-aspect>. Mode B model default suggestion: veo-3-1-fast (cheapest, fast). For best motion: kling-3. For abstract / artistic: fal-video with hosted model id.

This skill is distinct from:

  • reel-builder — full reels with script + music + captions, MP4 output
  • video-prompt — single-clip MP4 generation
  • subtitle-burner — captions on video (no animation)
  • bg-remover — image utility, not video

If the user wants to add captions to the GIF: not supported in v1. Workaround: use subtitle-burner --inline "<text>" on the intermediate MP4 (Mode B) BEFORE conversion, then GIF the result.

Signals

GitHub stars
21
Forks
1
Last commit
Sep 2026

ahel review

  • K1binfo
    installs-packages
  • K1binfo
    installs-packages (in references/troubleshoot.md)

Automated review, not a security audit. Ruleset v1+k2.

Advanced
Item type
skill
Key
gif-maker
Source
github.com/mikefluff/skills