Merge pull request #256 from MoonCaves/fix-gpt-image-default
replace retired openai image models with gpt-image-2 (dall-e 3 is shut down too)
This commit is contained in:
Calesthio
GitHub
commit
dc1cbca657
|
|
@ -16,7 +16,7 @@ GOOGLE_CLOUD_LOCATION= # Vertex AI region, default us-central1
|
|||
|
||||
# --- Voice ---
|
||||
ELEVENLABS_API_KEY= # TTS narration, music generation, sound effects
|
||||
OPENAI_API_KEY= # OpenAI TTS fallback and DALL-E image generation
|
||||
OPENAI_API_KEY= # OpenAI TTS fallback and GPT Image 2 image generation
|
||||
XAI_API_KEY= # Grok image generation/editing and Grok video generation
|
||||
DOUBAO_SPEECH_API_KEY= # Volcengine Doubao Speech TTS (new console API Key)
|
||||
DOUBAO_SPEECH_VOICE_TYPE= # Default Doubao speaker/voice type, e.g. zh_female_vv_uranus_bigtts
|
||||
|
|
|
|||
|
|
@ -489,7 +489,7 @@ Three selector tools abstract multi-provider capabilities. **Selectors auto-disc
|
|||
| Selector | Routes to | How it discovers |
|
||||
|----------|-----------|-----------------|
|
||||
| `tts_selector` | All tools with `capability="tts"` (ElevenLabs, Google TTS, OpenAI, Piper) | `registry.get_by_capability("tts")` |
|
||||
| `image_selector` | All tools with `capability="image_generation"` (FLUX, Google Imagen, DALL-E, Recraft, etc.) | `registry.get_by_capability("image_generation")` |
|
||||
| `image_selector` | All tools with `capability="image_generation"` (FLUX, Google Imagen, GPT Image, Recraft, etc.) | `registry.get_by_capability("image_generation")` |
|
||||
| `video_selector` | All tools with `capability="video_generation"` | `registry.get_by_capability("video_generation")` |
|
||||
|
||||
Selectors route based on: user preference > availability > discovery order. They adapt input schemas between providers transparently.
|
||||
|
|
|
|||
|
|
@ -183,7 +183,7 @@ SUNO_API_KEY=your-key # Full songs, instrumentals, any genre
|
|||
|
||||
# Voice & images:
|
||||
ELEVENLABS_API_KEY=your-key # Premium TTS, AI music, sound effects
|
||||
OPENAI_API_KEY=your-key # OpenAI TTS, DALL-E 3 images
|
||||
OPENAI_API_KEY=your-key # OpenAI TTS, GPT Image 2 images
|
||||
XAI_API_KEY=your-key # xAI Grok image edits/generation + Grok video generation
|
||||
GOOGLE_API_KEY=your-key # Google Imagen images, Google TTS (700+ voices)
|
||||
|
||||
|
|
@ -447,7 +447,7 @@ Each tool declares which Layer 3 skills it relies on. The agent reads Layer 1 to
|
|||
| **FLUX** | Cloud API | State-of-the-art quality |
|
||||
| **Google Imagen** | Cloud API | Imagen 4 — high-quality, multiple aspect ratios |
|
||||
| **Grok Imagine Image** | Cloud API | Strong image edits, style transfer, and multi-image compositing |
|
||||
| **DALL-E 3** | Cloud API | OpenAI's image model |
|
||||
| **GPT Image 2** | Cloud API | OpenAI's image model |
|
||||
| **Recraft** | Cloud API | Design-focused generation |
|
||||
| **Local Diffusion** | Local GPU | Stable Diffusion, free |
|
||||
| **Pexels** | Stock | Free stock images |
|
||||
|
|
|
|||
|
|
@ -177,7 +177,7 @@ SUNO_API_KEY=your-key # 完整的歌曲、伴奏，涵盖任何流派
|
|||
|
||||
# 语音与图像:
|
||||
ELEVENLABS_API_KEY=your-key # 顶级 TTS、AI 音乐、音效
|
||||
OPENAI_API_KEY=your-key # OpenAI TTS、DALL-E 3 图像
|
||||
OPENAI_API_KEY=your-key # OpenAI TTS、GPT Image 2 图像
|
||||
XAI_API_KEY=your-key # xAI Grok 图像编辑/生成 + Grok 视频生成
|
||||
GOOGLE_API_KEY=your-key # Google Imagen 图像、Google TTS（700+ 种声音）
|
||||
|
||||
|
|
@ -441,7 +441,7 @@ OpenMontage/
|
|||
| **FLUX** | 云端 API | 业界顶尖质量 |
|
||||
| **Google Imagen** | 云端 API | Imagen 4 — 高质量、多种长宽比 |
|
||||
| **Grok Imagine Image** | 云端 API | 强大的图像编辑、风格转换和多图合成 |
|
||||
| **DALL-E 3** | 云端 API | OpenAI 的图像模型 |
|
||||
| **GPT Image 2** | 云端 API | OpenAI 的图像模型 |
|
||||
| **Recraft** | 云端 API | 专注于设计的生成 |
|
||||
| **Local Diffusion** | 本地 GPU | Stable Diffusion，免费 |
|
||||
| **Pexels** | 素材库 | 免费的库存图片 |
|
||||
|
|
|
|||
|
|
@ -48,7 +48,7 @@ OpenMontage/
|
|||
│ ├── audio/ # TTS (ElevenLabs, OpenAI, Piper), music gen, mixing, enhancement
|
||||
│ ├── avatar/ # Talking head animation, lip sync
|
||||
│ ├── enhancement/ # Upscale, bg removal, face enhance/restore, color grading
|
||||
│ ├── graphics/ # Image gen (FLUX, DALL-E, Recraft, local diffusion), stock, diagrams, code snippets, math animation
|
||||
│ ├── graphics/ # Image gen (FLUX, GPT Image, Recraft, local diffusion), stock, diagrams, code snippets, math animation
|
||||
│ ├── publishers/ # (Reserved)
|
||||
│ ├── subtitle/ # SRT/VTT generation from timestamps
|
||||
│ └── video/ # 13 video gen providers, composition, stitching, trimming
|
||||
|
|
@ -383,7 +383,7 @@ All config is validated via Pydantic models in `lib/config_model.py`.
|
|||
| Variable | Used By | Purpose |
|
||||
|----------|---------|---------|
|
||||
| `ELEVENLABS_API_KEY` | elevenlabs_tts, music_gen | TTS, music, sound effects |
|
||||
| `OPENAI_API_KEY` | openai_tts, openai_image | TTS fallback, DALL-E 3 |
|
||||
| `OPENAI_API_KEY` | openai_tts, openai_image | TTS fallback, GPT Image 2 |
|
||||
| `XAI_API_KEY` | grok_image, grok_video | Grok image editing/generation, Grok video generation |
|
||||
| `FAL_KEY` | flux_image, kling_video, veo_video, minimax_video, recraft_image | fal.ai hosted models (FLUX, Veo, Kling, MiniMax, Recraft) |
|
||||
| `HEYGEN_API_KEY` | heygen_video | Multi-provider video generation |
|
||||
|
|
|
|||
|
|
@ -15,7 +15,7 @@ Everything you need to know about every provider in OpenMontage — setup instru
|
|||
| 3 | **$0** | ElevenLabs | Premium TTS + music + SFX (10K chars/month free) |
|
||||
| 4 | **$0** | Piper (local install) | Fully offline TTS — no API key, no cost, no network |
|
||||
| 5 | **~$0.03/image** | fal.ai | FLUX images + Kling/Veo/MiniMax video + Recraft — broad single-key image + video coverage |
|
||||
| 6 | **~$0.04/image** | OpenAI | DALL-E 3 images + OpenAI TTS |
|
||||
| 6 | **~$0.05/image** | OpenAI | GPT Image 2 images + OpenAI TTS |
|
||||
| 7 | **~$0.04/image** | Google Imagen | Imagen 4 images (shares the Google API key) |
|
||||
| 8 | **$12/month** | Runway | Gen-4 video — highest quality AI video |
|
||||
| 9 | **pay-as-you-go** | HeyGen | Avatar videos, multi-model video gateway |
|
||||
|
|
@ -37,7 +37,7 @@ GOOGLE_API_KEY= # Google TTS + Google Imagen
|
|||
|
||||
# VOICE + MUSIC
|
||||
ELEVENLABS_API_KEY= # TTS, music, sound effects (10K chars/month free)
|
||||
OPENAI_API_KEY= # OpenAI TTS + DALL-E 3 images
|
||||
OPENAI_API_KEY= # OpenAI TTS + GPT Image 2 images
|
||||
XAI_API_KEY= # xAI Grok image generation/editing + Grok video generation
|
||||
DOUBAO_SPEECH_API_KEY= # Volcengine Doubao Speech TTS (strong Mandarin narration)
|
||||
DOUBAO_SPEECH_VOICE_TYPE= # Default Doubao speaker/voice type
|
||||
|
|
@ -276,7 +276,7 @@ Google TTS offers 700+ voices across 50+ languages. Voice names follow the patte
|
|||
|
||||
### OpenAI — TTS + Image Generation
|
||||
|
||||
> **Solid all-rounder.** DALL-E 3 handles complex multi-element compositions well. TTS is fast and affordable.
|
||||
> **Solid all-rounder.** GPT Image 2 handles complex multi-element compositions and in-image text well. TTS is fast and affordable.
|
||||
|
||||
**Tools unlocked:** `openai_tts`, `openai_image`
|
||||
**Env var:** `OPENAI_API_KEY`
|
||||
|
|
@ -301,10 +301,14 @@ Google TTS offers 700+ voices across 50+ languages. Voice names follow the patte
|
|||
|
||||
| Model | Size | Quality | Price per image |
|
||||
|-------|------|---------|----------------|
|
||||
| DALL-E 3 | 1024x1024 | standard | $0.040 |
|
||||
| DALL-E 3 | 1024x1024 | hd | $0.080 |
|
||||
| DALL-E 3 | 1024x1792 | standard | $0.080 |
|
||||
| DALL-E 3 | 1024x1792 | hd | $0.120 |
|
||||
| GPT Image 2 | 1024x1024 | low | $0.006 |
|
||||
| GPT Image 2 | 1024x1024 | medium | $0.053 |
|
||||
| GPT Image 2 | 1024x1024 | high | $0.211 |
|
||||
| GPT Image 2 | 1024x1536 / 1536x1024 | low | $0.005 |
|
||||
| GPT Image 2 | 1024x1536 / 1536x1024 | medium | $0.041 |
|
||||
| GPT Image 2 | 1024x1536 / 1536x1024 | high | $0.165 |
|
||||
|
||||
> **Note:** DALL-E 2/3 were shut down by OpenAI on 2026-05-12, and the `gpt-image-1` family (`gpt-image-1-mini`, `gpt-image-1.5`) retires 2026-12-01 — `gpt-image-2` is OpenAI's recommended replacement ([deprecations](https://developers.openai.com/api/docs/deprecations)).
|
||||
|
||||
**Free tier:** None. Requires prepaid billing. Previously offered $5 in free credits for new accounts (discontinued for most signups).
|
||||
|
||||
|
|
@ -672,7 +676,7 @@ First run downloads the model (~4GB). Subsequent runs use the cached model.
|
|||
|
||||
**VRAM requirement:** 4GB+ (8GB recommended for 1024x1024 images)
|
||||
|
||||
**Supports:** Negative prompts, seeds, custom sizes. Quality is lower than FLUX or DALL-E 3 but completely free and offline.
|
||||
**Supports:** Negative prompts, seeds, custom sizes. Quality is lower than FLUX or GPT Image 2 but completely free and offline.
|
||||
|
||||
---
|
||||
|
||||
|
|
@ -743,7 +747,7 @@ How many providers cover each capability:
|
|||
|
||||
| Capability | Cloud Providers | Local Providers | Free Options |
|
||||
|-----------|----------------|-----------------|--------------|
|
||||
| **Image Generation** | FLUX, Grok, Google Imagen, DALL-E 3, Recraft | Local Diffusion | Pexels, Pixabay (stock) |
|
||||
| **Image Generation** | FLUX, Grok, Google Imagen, GPT Image 2, Recraft | Local Diffusion | Pexels, Pixabay (stock) |
|
||||
| **Video Generation** | Grok, Kling, Runway, Veo, Higgsfield, MiniMax, HeyGen | WAN, Hunyuan, CogVideo, LTX | Pexels, Pixabay (stock) |
|
||||
| **Text-to-Speech** | ElevenLabs, Google TTS, OpenAI | Piper | Piper, Google free tier, ElevenLabs free tier |
|
||||
| **Music Generation** | ElevenLabs, Suno | — | ElevenLabs free tier |
|
||||
|
|
|
|||
|
|
@ -12,7 +12,7 @@ clear, accurate, and effective in video.
|
|||
| Tool | Role |
|
||||
|------|------|
|
||||
| `diagram_gen` | Generate charts via Mermaid or D3 |
|
||||
| `image_selector` | Generate stylized chart illustrations (FLUX/DALL-E) |
|
||||
| `image_selector` | Generate stylized chart illustrations (FLUX/GPT Image) |
|
||||
| Remotion | Animated chart components (bar grow, line draw, pie fill) |
|
||||
| Manim | Mathematical plots, coordinate systems, function graphs |
|
||||
|
||||
|
|
|
|||
|
|
@ -1,6 +1,6 @@
|
|||
# Image Generation Usage for OpenMontage
|
||||
|
||||
> Sources: OpenAI DALL-E 3 documentation, FLUX/BFL API documentation, existing Layer 3 skills
|
||||
> Sources: OpenAI GPT Image documentation, FLUX/BFL API documentation, existing Layer 3 skills
|
||||
> at `.agents/skills/flux-best-practices/` and `.agents/skills/bfl-api/`
|
||||
|
||||
## Quick Reference Card
|
||||
|
|
@ -150,7 +150,7 @@ optimized for image/video generation providers.
|
|||
## Common Pitfalls
|
||||
|
||||
1. **Text in images** — AI image generators are unreliable with text. Never include text in prompts; add text as overlays in the compose stage
|
||||
2. **Hands and fingers** — DALL-E 3 and FLUX still struggle. Avoid prompts requiring detailed hand poses
|
||||
2. **Hands and fingers** — AI image models still struggle. Avoid prompts requiring detailed hand poses
|
||||
3. **Inconsistent characters** — Without reference images, the same character will look different each time. Always use the hero reference strategy
|
||||
4. **Over-prompting** — Long, complex prompts produce unpredictable results. Keep to 2-3 sentences
|
||||
5. **Over-unifying prompts** — Forcing the exact same style phrase into every prompt makes scenes look samey. Keep the visual system consistent, but let each scene express its own subject, shot, and emotional beat.
|
||||
|
|
|
|||
|
|
@ -11,7 +11,7 @@
|
|||
|------|----------|------|-------|----------|
|
||||
| `flux_image` | FLUX 2 Pro via fal.ai | ~$0.03-0.05 | ~5-10s | Photorealism, general purpose, workhorse |
|
||||
| `grok_image` | Grok Imagine Image (xAI) | $0.02/output + $0.002/input edit image | ~5-15s | Image edits, style transfer, multi-image compositing |
|
||||
| `openai_image` | GPT Image 1 (OpenAI) | ~$0.01-0.17 | ~5-15s | Complex instructions, text in images, multi-element |
|
||||
| `openai_image` | GPT Image 2 (OpenAI) | ~$0.01-0.21 | ~5-15s | Complex instructions, text in images, multi-element |
|
||||
| `recraft_image` | Recraft V4 via fal.ai | ~$0.04-0.25 | ~5-10s | Logos, SVG vectors, brand assets, text rendering (see caveat below) |
|
||||
| `local_diffusion` | Stable Diffusion (local) | Free | ~30s+ | Offline, privacy, free |
|
||||
| `image_gen` | Multi (legacy, deprecated) | Varies | Varies | **Deprecated** — use `image_selector` or per-provider tools |
|
||||
|
|
@ -39,7 +39,7 @@
|
|||
| **Style transfer / repaint of an existing image** | `grok_image` | Native edit flow, strong promptable transforms | `openai_image` |
|
||||
| **Multi-image merge / composite** | `grok_image` | Can combine multiple source images into one scene | `openai_image` |
|
||||
| **Logo or brand asset** | `recraft_image` | SVG support, text accuracy | `openai_image` |
|
||||
| **Image with text/labels** | `openai_image` | Best text rendering (GPT Image 1) | `recraft_image` |
|
||||
| **Image with text/labels** | `openai_image` | Best text rendering (GPT Image 2) | `recraft_image` |
|
||||
| **Complex multi-element composition** | `openai_image` | Best instruction following | `flux_image` |
|
||||
| **Hero image (key visual)** | `flux_image` | Highest visual quality | `openai_image` |
|
||||
| **Thumbnail** | `flux_image` or `recraft_image` | Needs to be eye-catching | — |
|
||||
|
|
@ -59,7 +59,7 @@
|
|||
PRODUCTION PATH: Premium
|
||||
├── Hero images: flux_image ($0.05/img)
|
||||
├── Supporting visuals: flux_image ($0.03/img)
|
||||
├── Text overlays: openai_image ($0.04/img)
|
||||
├── Text overlays: openai_image ($0.05/img medium)
|
||||
├── B-roll stills: pexels_image ($0.00)
|
||||
└── Total for 10 images: ~$0.35
|
||||
|
||||
|
|
|
|||
|
|
@ -159,7 +159,7 @@ If you encounter a generation technique, provider behavior, or prompting pattern
|
|||
|
||||
This is especially important for:
|
||||
- **Video generation prompting** — models respond to specific vocabularies that change with each version
|
||||
- **Image model parameters** — optimal settings for FLUX, DALL-E, Imagen differ and evolve
|
||||
- **Image model parameters** — optimal settings for FLUX, GPT Image, Imagen differ and evolve
|
||||
- **Audio provider quirks** — voice cloning, music generation, and TTS each have model-specific best practices
|
||||
- **Remotion component patterns** — new composition techniques emerge as the framework evolves
|
||||
|
||||
|
|
|
|||
|
|
@ -278,7 +278,7 @@ For each concept, specify:
|
|||
For each concept, specify:
|
||||
- **Animation approach**: `image_animation` / `clip_video` / `manim` / `remotion_dataviz` / `diagram_stills` / `mixed`
|
||||
- **Why this approach**: grounded in technique research AND tool availability from Step 3
|
||||
- **Image/video generation provider**: which specific provider from the preflight scan (e.g., "FLUX via fal.ai", "gpt-image-1 via OpenAI", "Stable Diffusion local")
|
||||
- **Image/video generation provider**: which specific provider from the preflight scan (e.g., "FLUX via fal.ai", "gpt-image-2 via OpenAI", "Stable Diffusion local")
|
||||
- **Reuse strategy**: What's the visual system? (recurring motifs, layout grid, color scheme, transition family)
|
||||
- **Complexity estimate**: How many unique scene types vs. reusable templates?
|
||||
- **Visual identity**: palette, typography, texture, motion energy, and why they fit this subject, audience, and platform
|
||||
|
|
@ -461,7 +461,7 @@ If you encounter a generation technique, provider behavior, or prompting pattern
|
|||
|
||||
This is especially important for:
|
||||
- **Video generation prompting** — models respond to specific vocabularies that change with each version
|
||||
- **Image model parameters** — optimal settings for FLUX, DALL-E, Imagen differ and evolve
|
||||
- **Image model parameters** — optimal settings for FLUX, GPT Image, Imagen differ and evolve
|
||||
- **Audio provider quirks** — voice cloning, music generation, and TTS each have model-specific best practices
|
||||
- **Remotion component patterns** — new composition techniques emerge as the framework evolves
|
||||
|
||||
|
|
|
|||
|
|
@ -124,7 +124,7 @@ If you encounter a generation technique, provider behavior, or prompting pattern
|
|||
|
||||
This is especially important for:
|
||||
- **Video generation prompting** — models respond to specific vocabularies that change with each version
|
||||
- **Image model parameters** — optimal settings for FLUX, DALL-E, Imagen differ and evolve
|
||||
- **Image model parameters** — optimal settings for FLUX, GPT Image, Imagen differ and evolve
|
||||
- **Audio provider quirks** — voice cloning, music generation, and TTS each have model-specific best practices
|
||||
- **Remotion component patterns** — new composition techniques emerge as the framework evolves
|
||||
|
||||
|
|
|
|||
|
|
@ -157,7 +157,7 @@ If you encounter a generation technique, provider behavior, or prompting pattern
|
|||
|
||||
This is especially important for:
|
||||
- **Video generation prompting** — models respond to specific vocabularies that change with each version
|
||||
- **Image model parameters** — optimal settings for FLUX, DALL-E, Imagen differ and evolve
|
||||
- **Image model parameters** — optimal settings for FLUX, GPT Image, Imagen differ and evolve
|
||||
- **Audio provider quirks** — voice cloning, music generation, and TTS each have model-specific best practices
|
||||
- **Remotion component patterns** — new composition techniques emerge as the framework evolves
|
||||
|
||||
|
|
|
|||
|
|
@ -285,7 +285,7 @@ If you encounter a generation technique, provider behavior, or prompting pattern
|
|||
|
||||
This is especially important for:
|
||||
- **Video generation prompting** — models respond to specific vocabularies that change with each version
|
||||
- **Image model parameters** — optimal settings for FLUX, DALL-E, Imagen differ and evolve
|
||||
- **Image model parameters** — optimal settings for FLUX, GPT Image, Imagen differ and evolve
|
||||
- **Audio provider quirks** — voice cloning, music generation, and TTS each have model-specific best practices
|
||||
- **Remotion component patterns** — new composition techniques emerge as the framework evolves
|
||||
|
||||
|
|
|
|||
|
|
@ -101,7 +101,7 @@ If you encounter a generation technique, provider behavior, or prompting pattern
|
|||
|
||||
This is especially important for:
|
||||
- **Video generation prompting** — models respond to specific vocabularies that change with each version
|
||||
- **Image model parameters** — optimal settings for FLUX, DALL-E, Imagen differ and evolve
|
||||
- **Image model parameters** — optimal settings for FLUX, GPT Image, Imagen differ and evolve
|
||||
- **Audio provider quirks** — voice cloning, music generation, and TTS each have model-specific best practices
|
||||
- **Remotion component patterns** — new composition techniques emerge as the framework evolves
|
||||
|
||||
|
|
|
|||
|
|
@ -273,7 +273,7 @@ If you encounter a generation technique, provider behavior, or prompting pattern
|
|||
|
||||
This is especially important for:
|
||||
- **Video generation prompting** — models respond to specific vocabularies that change with each version
|
||||
- **Image model parameters** — optimal settings for FLUX, DALL-E, Imagen differ and evolve
|
||||
- **Image model parameters** — optimal settings for FLUX, GPT Image, Imagen differ and evolve
|
||||
- **Audio provider quirks** — voice cloning, music generation, and TTS each have model-specific best practices
|
||||
- **Remotion component patterns** — new composition techniques emerge as the framework evolves
|
||||
|
||||
|
|
|
|||
|
|
@ -535,7 +535,7 @@ If you encounter a generation technique, provider behavior, or prompting pattern
|
|||
|
||||
This is especially important for:
|
||||
- **Video generation prompting** — models respond to specific vocabularies that change with each version
|
||||
- **Image model parameters** — optimal settings for FLUX, DALL-E, Imagen differ and evolve
|
||||
- **Image model parameters** — optimal settings for FLUX, GPT Image, Imagen differ and evolve
|
||||
- **Audio provider quirks** — voice cloning, music generation, and TTS each have model-specific best practices
|
||||
- **Remotion component patterns** — new composition techniques emerge as the framework evolves
|
||||
|
||||
|
|
|
|||
|
|
@ -79,7 +79,7 @@ Transform each script section into 1-3 visual scenes. Each scene is a distinct v
|
|||
| `text_card` | Statements, closing messages, key terms | Remotion TextCard (centered, spring animation) | 3-5s |
|
||||
| `animation` | Concepts needing motion (data flow, math) | Remotion, Manim | 4-10s |
|
||||
| `diagram` | Processes, architecture, relationships | `diagram_gen` (Mermaid), `image_selector` | 4-8s |
|
||||
| `generated` | Illustrations, metaphors, real-world imagery | `image_selector` (FLUX/DALL-E) | 3-6s |
|
||||
| `generated` | Illustrations, metaphors, real-world imagery | `image_selector` (FLUX/GPT Image) | 3-6s |
|
||||
| `talking_head` | AI avatar speaking (if HeyGen available) | HeyGen tools | 5-15s |
|
||||
| `broll` | Context, real-world examples | Stock or generated footage | 3-6s |
|
||||
| `screen_recording` | Code demos, UI walkthroughs | Recorded or simulated | 5-15s |
|
||||
|
|
@ -206,7 +206,7 @@ The style playbook constrains your visual choices:
|
|||
**Feasibility check:**
|
||||
- [ ] Every `required_asset` with `source: "generate"` is achievable with available tools
|
||||
- [ ] Diagram descriptions are specific enough for Mermaid syntax generation
|
||||
- [ ] Image descriptions are specific enough for FLUX/DALL-E prompt engineering
|
||||
- [ ] Image descriptions are specific enough for FLUX/GPT Image prompt engineering
|
||||
- [ ] No scene requires tools that aren't in the tool registry
|
||||
|
||||
### Step 7: Self-Evaluate
|
||||
|
|
|
|||
|
|
@ -93,7 +93,7 @@ If you encounter a generation technique, provider behavior, or prompting pattern
|
|||
|
||||
This is especially important for:
|
||||
- **Video generation prompting** — models respond to specific vocabularies that change with each version
|
||||
- **Image model parameters** — optimal settings for FLUX, DALL-E, Imagen differ and evolve
|
||||
- **Image model parameters** — optimal settings for FLUX, GPT Image, Imagen differ and evolve
|
||||
- **Audio provider quirks** — voice cloning, music generation, and TTS each have model-specific best practices
|
||||
- **Remotion component patterns** — new composition techniques emerge as the framework evolves
|
||||
|
||||
|
|
|
|||
|
|
@ -85,7 +85,7 @@ If you encounter a generation technique, provider behavior, or prompting pattern
|
|||
|
||||
This is especially important for:
|
||||
- **Video generation prompting** — models respond to specific vocabularies that change with each version
|
||||
- **Image model parameters** — optimal settings for FLUX, DALL-E, Imagen differ and evolve
|
||||
- **Image model parameters** — optimal settings for FLUX, GPT Image, Imagen differ and evolve
|
||||
- **Audio provider quirks** — voice cloning, music generation, and TTS each have model-specific best practices
|
||||
- **Remotion component patterns** — new composition techniques emerge as the framework evolves
|
||||
|
||||
|
|
|
|||
|
|
@ -98,7 +98,7 @@ If you encounter a generation technique, provider behavior, or prompting pattern
|
|||
|
||||
This is especially important for:
|
||||
- **Video generation prompting** — models respond to specific vocabularies that change with each version
|
||||
- **Image model parameters** — optimal settings for FLUX, DALL-E, Imagen differ and evolve
|
||||
- **Image model parameters** — optimal settings for FLUX, GPT Image, Imagen differ and evolve
|
||||
- **Audio provider quirks** — voice cloning, music generation, and TTS each have model-specific best practices
|
||||
- **Remotion component patterns** — new composition techniques emerge as the framework evolves
|
||||
|
||||
|
|
|
|||
|
|
@ -159,7 +159,7 @@ If you encounter a generation technique, provider behavior, or prompting pattern
|
|||
|
||||
This is especially important for:
|
||||
- **Video generation prompting** — models respond to specific vocabularies that change with each version
|
||||
- **Image model parameters** — optimal settings for FLUX, DALL-E, Imagen differ and evolve
|
||||
- **Image model parameters** — optimal settings for FLUX, GPT Image, Imagen differ and evolve
|
||||
- **Audio provider quirks** — voice cloning, music generation, and TTS each have model-specific best practices
|
||||
- **Remotion component patterns** — new composition techniques emerge as the framework evolves
|
||||
|
||||
|
|
|
|||
|
|
@ -201,7 +201,7 @@ If you encounter a generation technique, provider behavior, or prompting pattern
|
|||
|
||||
This is especially important for:
|
||||
- **Video generation prompting** — models respond to specific vocabularies that change with each version
|
||||
- **Image model parameters** — optimal settings for FLUX, DALL-E, Imagen differ and evolve
|
||||
- **Image model parameters** — optimal settings for FLUX, GPT Image, Imagen differ and evolve
|
||||
- **Audio provider quirks** — voice cloning, music generation, and TTS each have model-specific best practices
|
||||
- **Remotion component patterns** — new composition techniques emerge as the framework evolves
|
||||
|
||||
|
|
|
|||
|
|
@ -9,7 +9,7 @@ Run every tool with real API keys, inspect outputs (see images, listen to audio,
|
|||
| Script | Tools Tested | API Keys Used | Est. Cost |
|
||||
|--------|-------------|---------------|-----------|
|
||||
| `test_01_tts.py` | `elevenlabs_tts` (ElevenLabs) | ELEVENLABS_API_KEY | ~$0.02 |
|
||||
| `test_02_image_gen.py` | `image_gen` (DALL-E 3 + FLUX) | OPENAI_API_KEY, FAL_AI_API_KEY | ~$0.15 |
|
||||
| `test_02_image_gen.py` | `image_gen` (GPT Image 2 + FLUX) | OPENAI_API_KEY, FAL_AI_API_KEY | ~$0.15 |
|
||||
| `test_03_music.py` | `music_gen` (ElevenLabs) | ELEVENLABS_API_KEY | ~$0.10 |
|
||||
| `test_04_audio_mix.py` | `audio_mixer` | None (ffmpeg only) | $0 |
|
||||
| `test_05_video_compose.py` | `video_compose` | None (ffmpeg only) | $0 |
|
||||
|
|
@ -30,7 +30,7 @@ For each output:
|
|||
| Area | Risk | How to Validate |
|
||||
|------|------|-----------------|
|
||||
| TTS voice selection | Default voice may not match playbook mood | Test with multiple voice IDs, compare against playbook `voice_style` |
|
||||
| Image gen consistency | DALL-E/FLUX outputs vary wildly per prompt | Test with playbook `image_prompt_prefix` prepended |
|
||||
| Image gen consistency | GPT Image/FLUX outputs vary wildly per prompt | Test with playbook `image_prompt_prefix` prepended |
|
||||
| Music duration alignment | Music may not match narration duration | Compare `music.duration` vs `tts.duration`, check padding/looping |
|
||||
| Audio ducking timing | Ducking may cut music too aggressively | Inspect waveform: music should duck ~6dB under speech, recover smoothly |
|
||||
| Video stitch transitions | Crossfade may flicker with mismatched codecs | Test with both matching and mismatched clips, check `auto_normalize` |
|
||||
|
|
|
|||
|
|
@ -7,7 +7,7 @@
|
|||
pexels_image, pixabay_image). This file is kept for backwards
|
||||
compatibility and will be removed in a future release.
|
||||
|
||||
Supports cloud API providers (FLUX via fal.ai/Replicate, OpenAI DALL-E)
|
||||
Supports cloud API providers (FLUX via fal.ai/Replicate, OpenAI GPT Image)
|
||||
and local Stable Diffusion via diffusers. Reports unavailable with
|
||||
install instructions when no provider is configured.
|
||||
"""
|
||||
|
|
@ -43,12 +43,12 @@ class ImageGen(BaseTool):
|
|||
stability = ToolStability.EXPERIMENTAL
|
||||
execution_mode = ExecutionMode.SYNC
|
||||
determinism = Determinism.SEEDED
|
||||
runtime = ToolRuntime.HYBRID # API (DALL-E/FLUX) or local (diffusers)
|
||||
runtime = ToolRuntime.HYBRID # API (GPT Image/FLUX) or local (diffusers)
|
||||
|
||||
dependencies = [] # checked dynamically based on provider
|
||||
install_instructions = (
|
||||
"Set one of these environment variables:\n"
|
||||
" OPENAI_API_KEY — for DALL-E 3\n"
|
||||
" OPENAI_API_KEY — for GPT Image 2\n"
|
||||
" FAL_KEY — for FLUX via fal.ai\n"
|
||||
"Or install diffusers for local generation:\n"
|
||||
" pip install diffusers transformers accelerate torch"
|
||||
|
|
@ -121,7 +121,7 @@ class ImageGen(BaseTool):
|
|||
def estimate_cost(self, inputs: dict[str, Any]) -> float:
|
||||
provider = inputs.get("provider") or self._detect_provider()
|
||||
if provider == "openai":
|
||||
return 0.04 # DALL-E 3 standard
|
||||
return 0.053 # gpt-image-2 medium at 1024x1024 (call uses auto quality)
|
||||
if provider == "flux":
|
||||
return 0.03
|
||||
return 0.0 # local
|
||||
|
|
@ -159,14 +159,14 @@ class ImageGen(BaseTool):
|
|||
client = OpenAI()
|
||||
prompt = inputs["prompt"]
|
||||
size = f"{inputs.get('width', 1024)}x{inputs.get('height', 1024)}"
|
||||
model = inputs.get("model", "dall-e-3")
|
||||
model = inputs.get("model", "gpt-image-2")
|
||||
|
||||
# GPT image models don't accept response_format; they always return b64
|
||||
response = client.images.generate(
|
||||
model=model,
|
||||
prompt=prompt,
|
||||
size=size,
|
||||
n=1,
|
||||
response_format="b64_json",
|
||||
)
|
||||
|
||||
image_data = base64.b64decode(response.data[0].b64_json)
|
||||
|
|
|
|||
|
|
@ -1,4 +1,4 @@
|
|||
"""OpenAI GPT Image generation (gpt-image-1 / DALL-E 3)."""
|
||||
"""OpenAI GPT Image generation (gpt-image-2)."""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
|
|
@ -60,20 +60,17 @@ class OpenAIImage(BaseTool):
|
|||
"prompt": {"type": "string"},
|
||||
"model": {
|
||||
"type": "string",
|
||||
"enum": ["gpt-image-1", "dall-e-3"],
|
||||
"default": "gpt-image-1",
|
||||
"enum": ["gpt-image-2"],
|
||||
"default": "gpt-image-2",
|
||||
},
|
||||
"size": {
|
||||
"type": "string",
|
||||
"enum": [
|
||||
"1024x1024", "1536x1024", "1024x1536", "auto",
|
||||
"1024x1792", "1792x1024", # dall-e-3 only
|
||||
],
|
||||
"enum": ["1024x1024", "1536x1024", "1024x1536", "auto"],
|
||||
"default": "1024x1024",
|
||||
},
|
||||
"quality": {
|
||||
"type": "string",
|
||||
"enum": ["low", "medium", "high", "auto", "standard", "hd"],
|
||||
"enum": ["low", "medium", "high", "auto"],
|
||||
"default": "high",
|
||||
},
|
||||
"output_format": {
|
||||
|
|
@ -100,15 +97,12 @@ class OpenAIImage(BaseTool):
|
|||
return ToolStatus.UNAVAILABLE
|
||||
|
||||
def estimate_cost(self, inputs: dict[str, Any]) -> float:
|
||||
model = inputs.get("model", "gpt-image-1")
|
||||
# gpt-image-2 per-image pricing at 1024x1024 (non-square sizes run
|
||||
# slightly cheaper): https://developers.openai.com/api/docs/guides/image-generation
|
||||
quality = inputs.get("quality", "high")
|
||||
n = inputs.get("n", 1)
|
||||
if model == "gpt-image-1":
|
||||
cost_map = {"low": 0.011, "medium": 0.042, "high": 0.167, "auto": 0.042}
|
||||
return cost_map.get(quality, 0.042) * n
|
||||
# dall-e-3 fallback pricing
|
||||
quality_map = {"standard": 0.04, "hd": 0.08}
|
||||
return quality_map.get(quality, 0.04) * n
|
||||
cost_map = {"low": 0.006, "medium": 0.053, "high": 0.211, "auto": 0.053}
|
||||
return cost_map.get(quality, 0.053) * n
|
||||
|
||||
def execute(self, inputs: dict[str, Any]) -> ToolResult:
|
||||
if not os.environ.get("OPENAI_API_KEY"):
|
||||
|
|
@ -121,36 +115,22 @@ class OpenAIImage(BaseTool):
|
|||
|
||||
start = time.time()
|
||||
client = OpenAI()
|
||||
model = inputs.get("model", "gpt-image-1")
|
||||
model = inputs.get("model", "gpt-image-2")
|
||||
prompt = inputs["prompt"]
|
||||
size = inputs.get("size", "1024x1024")
|
||||
n = inputs.get("n", 1)
|
||||
|
||||
try:
|
||||
if model == "gpt-image-1":
|
||||
quality = inputs.get("quality", "high")
|
||||
output_format = inputs.get("output_format", "png")
|
||||
response = client.images.generate(
|
||||
model=model,
|
||||
prompt=prompt,
|
||||
size=size,
|
||||
quality=quality,
|
||||
output_format=output_format,
|
||||
n=n,
|
||||
)
|
||||
else:
|
||||
# dall-e-3 path
|
||||
quality = inputs.get("quality", "standard")
|
||||
if quality in ("low", "medium", "high", "auto"):
|
||||
quality = "standard" # map to dall-e-3 quality options
|
||||
response = client.images.generate(
|
||||
model=model,
|
||||
prompt=prompt,
|
||||
size=size,
|
||||
quality=quality,
|
||||
n=1, # dall-e-3 only supports n=1
|
||||
response_format="b64_json",
|
||||
)
|
||||
quality = inputs.get("quality", "high")
|
||||
output_format = inputs.get("output_format", "png")
|
||||
response = client.images.generate(
|
||||
model=model,
|
||||
prompt=prompt,
|
||||
size=size,
|
||||
quality=quality,
|
||||
output_format=output_format,
|
||||
n=n,
|
||||
)
|
||||
|
||||
image_data = base64.b64decode(response.data[0].b64_json)
|
||||
ext = inputs.get("output_format", "png")
|
||||
|
|
|
|||
Loading…
Reference in New Issue