ChatGPT
General AI assistant with Images 2.5 stills, an agent for multi-step work, and a Word sidebar
ChatGPT is OpenAI's general assistant for writing, research, coding and image work. In 2026 it gained ChatGPT Images 2.5 for stills, the ChatGPT Work agent (launched July 2026) for multi-step document and deck tasks, and a sidebar inside Microsoft Word. Video generation was removed when OpenAI shut down Sora in 2026.
- 223 creators using
- Freemium pricing
- Easy learning curve
What ChatGPT can do
- Infinite canvas No
- AI agent Yes
- Node / flow builder No
- Public API Yes
- MCP server Partial
13 of 26 capabilities confirmed
| Capability | Supported | Details |
|---|---|---|
| Text to image | Yes | ChatGPT Images 2.5 (Sept 2026); thinking mode on paid plans. |
| Image editing | Yes | In-image editing and comments in Images 2.5. |
| Text to video | No | Sora removed; no video generation in ChatGPT. |
| Image to video | No | |
| Video to video | No | |
| Native audio in video | No | |
| Lip-sync | No | |
| Voice / TTS | Partial | Voice conversations with plugins on web, iOS and Android (Sept 2026). Conversational speech only; no custom voice cloning or TTS export found. |
| Music & SFX | No | |
| Upscaling | Not confirmed | Not verified. |
| Character consistency | Partial | Images 2.5 keeps reference-photo subjects more accurate and holds edits across long conversations. No persistent character training. |
| Camera control | Not confirmed | Not verified for image work. |
| Custom training | No | Custom GPTs and workspace agents are prompt-and-tool configurations, not model training. |
| Third-party models | No | Offers OpenAI models only. |
| Capability | Supported | Details |
|---|---|---|
| Infinite canvas | No | Standalone Canvas side panel reportedly removed from main models on 28 May 2026; writing and code now sit inline as blocks. Not an infinite board. |
| Node / flow builder | No | |
| AI agent | Yes | ChatGPT Work (launched 9 July 2026): multi-step document, spreadsheet, deck and web-app tasks across connected apps and files. Free and Go tiers excluded. |
| Chat assistant | Yes | Core product. |
| Team collaboration | Yes | Business and Enterprise workspaces; workspace agents replace custom GPTs for teams. |
| Capability | Supported | Details |
|---|---|---|
| Public API | Yes | OpenAI API: Responses API, Agents SDK, gpt-image-2.5 models. |
| MCP server | Partial | ChatGPT accepts custom remote MCP servers as connectors. Apps SDK (MCP-based ChatGPT apps) is listed as preview in OpenAI docs. ChatGPT is a client, not a published MCP server. |
| Mobile app | Yes | iOS and Android. |
| Desktop / local | Partial | Desktop app exists; models run in OpenAI's cloud, not locally. |
| Open source | No |
| Capability | Supported | Details |
|---|---|---|
| Free tier | Yes | Free plan with limited messages and uploads; reported to carry ads in the US. |
| Commercial use | Yes | Paid plans carry commercial rights; Business and Enterprise data excluded from training by default. |
Yes Partial No Not confirmed by the vendor
What makes ChatGPT different
- 01
Stills with reasoning: the Images 2.x line plans the composition and checks object counts and prompt constraints before drawing, and renders legible text, which helps with posters, title cards and storyboard labels.
- 02
Agent that produces files: ChatGPT Work outputs real .docx, spreadsheet, slide and small web app deliverables from connected apps, not just chat answers.
- 03
Widest general-purpose surface: one subscription covers scripting, research, image stills, voice and file analysis, which most single-purpose AI video tools do not cover.
- 04
Remote MCP and API access: developers can point the OpenAI Responses API at any remote MCP server, and ChatGPT accepts custom MCP connectors.
- 05
No in-app video: unlike Gemini (Veo through Flow) or Higgsfield and Kling, ChatGPT is not a shot generator, so it does not compete for the video budget.
About ChatGPT
ChatGPT is still the most widely used AI assistant, and for filmmakers its main value is text and stills, not moving images. Image generation runs on ChatGPT Images 2.5 (released 8 September 2026), which follows GPT Image 2 (21 April 2026). The 2.5 release adds sketch-to-image input, templates for posters and flyers, comments on specific parts of an image, and more consistent multi-turn edits. Thinking mode, a planning pass before drawing, is limited to paid plans.
The bigger 2026 change is ChatGPT Work, an agent launched 9 July 2026 with the GPT-5.6 family (Sol, Terra, Luna). It gathers context across connected apps and files and produces documents, spreadsheets, presentations and small web apps. Its Data plugin (September 2026) analyses connected business data and builds dashboards. The Word sidebar (September 2026) drafts from notes and revises selected text on all plans, including Free.
Video is the big gap. OpenAI closed the Sora app on 26 April 2026 and reportedly closed the Sora API on 24 September 2026, and ChatGPT no longer generates video. For shot generation, a filmmaker needs a separate video tool. ChatGPT fits pre-production: scripts, shot lists, treatments, research, prompt writing, and stills for look development.
What's new in ChatGPT
-
ChatGPT Images 2.5 launches
Sharper detail, faster generation (up to 50% lower latency than 2.0), Sketch input, Templates, in-image comments and more consistent multi-turn edits.
Source: openai.com ↗ -
ChatGPT sidebar in Microsoft Word
Drafts from notes, summarizes documents and revises selected text; available on all plans including Free.
Source: help.openai.com ↗ -
Data plugin for ChatGPT Work
Analyses connected business data and builds dashboards or reports from team metric definitions.
Source: help.openai.com ↗ -
ChatGPT Work agent and GPT-5.6 family
Agent that works across apps and files to produce documents, spreadsheets, presentations and web apps; model tiers Sol, Terra and Luna.
Source: openai.com ↗ -
Sora app and API shutdown begins
The Sora app closed 26 April 2026; the API was scheduled to close 24 September 2026. Video generation is no longer offered in ChatGPT.
Source: the-decoder.com ↗
Models available in ChatGPT
-
GPT Image 2 OpenAI April 2026 release; up to 2K output, up to 8 images per prompt, thinking mode on paid plans.
-
-
Key Features
- ChatGPT Images 2.5: stills with sketch input, templates, in-image comments and prompt sharing
- ChatGPT Work agent: documents, spreadsheets, decks and web apps from connected apps and files
- ChatGPT for Word sidebar: draft, summarize and revise selected text
- Data plugin: analyse connected business data, build dashboards
- Deep research on paid plans
- Voice conversations with plugins on web and mobile
- File uploads and analysis (Python data analysis)
- Custom GPTs and workspace agents
- Custom remote MCP servers as connectors
- GPT-5.6 Sol, Terra and Luna model selection on paid plans
- Codex coding agent included on paid plans
When to reach for it — and when to skip
Reach for it when…
- Strong text-in-image rendering for posters, title cards and slate labels
- Agent output is real files (decks, sheets, docs), not only chat text
- One subscription covers scripting, research, stills and voice
- Mature API and MCP connector support for building custom pipelines
- Commercial use on paid plans with training opt-out for business data
Skip it when…
- No video generation after the Sora shutdown; shot generation needs another tool
- Plan limits are dynamic and hard to predict, especially for agent runs
- Free tier carries ads (reported) and tight message caps
- Model names and tiers shift often (GPT-5.5, 5.6 and unconfirmed GPT-6 reports), making results hard to pin down
- Canvas side panel removed from main models, so long document editing is less convenient
Best For
✓ Ideal for
- Writing shot lists, treatments, pitch decks and scripts with file output
- Generating look-development stills and poster or title-card mockups with readable text
- Analysing budget, schedule or audience spreadsheets with the Data plugin
- Drafting and revising documents directly in Word on a client's machine
- Building custom tools via the API or remote MCP connectors
✗ Not built for
- Generating video clips or shots (use a dedicated video model tool)
- Persistent character or identity training for a production (use a LoRA or Soul-style tool)
- Node-based pipelines or infinite canvas boards (use ComfyUI, Krea or Flora)
Working Tips from Filmmakers Using ChatGPT
- 01 Use ChatGPT for the prompt and shot list, then render the clips in a video tool; it no longer makes video.
- 02 Images 2.5 Sketch input works well for rough blocking: draw a crude frame, describe the lighting, and iterate with image comments.
- 03 Agent runs burn allowance fast; batch related requests into one brief rather than many small ones.
- 04 Keep sensitive client scripts out of Free and Go plans; Business and Enterprise data is excluded from training by default.
ChatGPT pricing in 2026
| Plan | Price | What you get |
|---|---|---|
| Free | $0 monthly |
|
| Go | $8 monthly |
|
| Plus | $20 monthly |
|
| Pro (Pro Lite) | $100 monthly |
|
| Pro | $200 monthly |
|
| Business | $20 per user (annual) / $25 (monthly) per seat |
|
| Enterprise | Custom annual |
|
Prices checked against ChatGPT's pricing page on 9 Oct 2026. Vendors change plans often, so confirm before you buy.
ChatGPT vs Claude vs Gemini
| ChatGPT | Claude | Gemini | |
|---|---|---|---|
| Price | Free (ads, US) / Go $8/mo / Plus $20/mo / Pro $100 or $200/mo; Business about $20-25 per user/mo; Enterprise custom | Free; Pro $20/mo ($17/mo billed annually); Max from $100/mo | Free; Google AI Plus from €4.99/mo; AI Pro €21.99/mo; AI Ultra from €99.99/mo (EUR) |
| Infinite canvas | No | Partial | Yes |
| AI agent | Yes | Yes | Partial |
| Chat assistant | Yes | Yes | Yes |
| Team collaboration | Yes | Yes | Partial |
| Text to image | Yes | No | Yes |
| Image editing | Yes | No | Yes |
| Text to video | No | No | Yes |
| Image to video | No | No | Yes |
| Video to video | No | No | Yes |
| Native audio in video | No | No | Yes |
| Voice / TTS | Partial | No | Not confirmed |
| Upscaling | Not confirmed | No | Yes |
| Character consistency | Partial | No | Partial |
| Camera control | Not confirmed | No | Partial |
| Public API | Yes | Yes | Yes |
| MCP server | Partial | Partial | Partial |
| Mobile app | Yes | Yes | Yes |
| Desktop / local | Partial | Yes | No |
| Free tier | Yes | Yes | Yes |
| Commercial use | Yes | Yes | Not confirmed |
| Entry paid plan | Plus $20/mo | Pro $20/mo | AI Pro (EUR 21.99/mo on DE page) |
| In-app video generation | No (Sora removed) | Team/Enterprise only (Claude Motion, beta) | Yes (Veo via Flow; Pro and Ultra credits) |
| In-app image generation | Yes (Images 2.5) | Not listed on pricing page | Yes (Nano Banana via Flow and Search) |
| Agent mode | Yes (ChatGPT Work) | Partial (Claude Code; scheduled tasks on Pro and Max) | Partial (Deep Research; Antigravity on paid plans) |
| Voice | Yes | Yes | Yes (Gemini Live) |
| Public API | Yes (OpenAI API, Responses API, MCP) | Yes (usage billed at API rates) | Not checked |
Each column comes from that tool's own researched page. "Not confirmed" means the vendor doesn't say.
Open the full ChatGPT vs Claude vs Gemini comparison →Use Cases
Tags
Sources
Facts on this page were checked against ChatGPT's own pricing, product and changelog pages on , then cross-checked with independent coverage where vendor pages were silent. Anything we couldn't confirm is marked "Not confirmed" rather than guessed.
- ChatGPT pricing page chatgpt.com
- ChatGPT release notes (OpenAI help center; surfaced via search, direct fetch returned 403) help.openai.com
- Introducing ChatGPT Images 2.5 (OpenAI, 8 Sept 2026; surfaced via search) openai.com
- GPT-5.6 announcement (OpenAI; surfaced via search, direct fetch returned 403) openai.com
- OpenAI docs: remote MCP servers developers.openai.com
- The Decoder: two-stage Sora shutdown the-decoder.com
- NeuroHive: ChatGPT Images 2.0 launch neurohive.io
- CometAPI: ChatGPT pricing 2026 (third party) cometapi.com
- Claude pricing page (comparison column) claude.com
- Google AI plans page, German edition in EUR (comparison column) gemini.google
Spotted something out of date? Tell us and we'll re-check it.
Discussion
No comments yet — be the first.